computation of spatially varying ground ...beskos [28] and sanchez-sesma [29]. quasi two-dimensional...
TRANSCRIPT
REPORT NO.
UCB/EERC-91/06
MAY 1991
PB93-114825
EARTHQUAKE ENGINEERING RESEARCH CENTER
COMPUTATION OFSPATIAllY VARYING GROUND MOTION ANDFOUNDATION-ROCK IMPEDANCE MATRICESFOR SEISMIC ANALYSIS OF ARCH DAMS
by
llPING ZHANG
ANIL K. CHOPRA
A Report on Research conducted underGrant No. CES-8719296 from the
, ...,.~NationalScience Foundation
COLLEGE OF ENGINEERING
UNIVERSITY OF CALIFORNIA AT BERKELEY
For sale by the National Technical Information
Service, U.S. Department of Commerce, Spring
field, Virginia 22161
See back of report for up to date listing of EERC
reports.
DISCLAIMER
Any opinions, findings, and conclusions or
recommendations expressed in this publication
are those of the authors and do not necessarilyreflect the views of the National Science Foun
dation or the Earthquake Engineering Research
Center, University of Califomia at Berkeley.
COMPUTATION OF SPATIALLY VARYING GROUND MOTION
AND FOUNDATION-ROCK IMPEDANCE MATRICES
FOR SEISMIC ANALYSIS OF ARCH DAMS
by
Liping Zhang
Anil K. Chopra
A Report on Research Conducted underGrant CES-8719296 from
the National Science Foundation
Report No. UCB/EERC 91/06Earthquake Engineering Research Center
University of CaliforniaBerkeley, California
May 1991
ABSTRACT'
This work on the development of analytical procedures to solve two problems
that are motivated by the earthquake analysis of arch dams is organized in two parts.
Presented in the first part is a direct boundary element method (BEM) to deter
mine the three-dimensional seismic response of an infinitely-long canyon of arbitrary
but uniform cross-section cut in a homogeneous viscoelastic half-space. The seismic
excitation is represented by P, SV, SH or Rayleigh waves at arbitrary angles with
respect to the axis of the canyon. The accuracy of the procedure and implementing
computer program is demonstrated by comparison with previous solutions for the li
miting case of two-dimensional response, recently obtained three-dimensional response
results for infinitely-long canyons, and three-dimensional boundary element method
solutions presented in this work for finite canyons. This procedure would enable ana
lytical estimation of the spatial variation of ground motions around the canyon and
hence the possibility of exploring the effects of such variation on earthquake response
of arch dams.
In the second part, a direct boundary element procedure is presented to determine
the foundation impedance matrix defined at the nodal points on the dam-foundation
rock interface. The uniform cross-section of the infinitely-long canyon permits analy
tical integration along the canyon axis leading to a series of two-dimensional boundary
problems involving Fourier transforms of the full-space Green's functions. Solution
of these two-dimensional boundary problems leads to a dynamic flexibility influence
matrix which is inverted to determine the impedance matrix. The accuracy of the pro
cedure is demonstrated by comparison with previous solutions for a surface-supported,
square foundation and results obtained by a three-dimensional BEM for a foundation
of finite-width supported on an infinitely-long canyon. Compared with the three
dimensional BEM, the present method requires less computer storage and is more
accurate and efficient. The foundation impedance matrix determined by this proce
dure can be incorporated in a substructure method for earthquake analysis of arch
dams.
ACKNOWLEDGEMENTS
This research investigation was supported by the National Science Foundation
under Grant CES-8719296. The authors are grateful for this support.
Parts of this work were accomplished during the 1990 Fall semester while Anil
K. Chopra was on appointment as a Miller Research Professor in the Miller Institute
of Basic Research in Science, University of California at Berkeley.
ii
TABLE OF CONTENTS
ABSTRACT 1
ACKNOWLEDGEMENTS ii
TABLE OF CONTENTS iii
PART I: THREE-DIMENSIONAL ANALYSIS OF SPATIALLY VARY
ING GROUND MOTIONS AROUND A UNIFORM CANYON IN A
HOMOGENEOUS HALF-SPACE , 1
INTRODUCTION . 3
SYSTEM CONSIDERED 6
FREE FIELD MOTION 8
BOUNDARY INTEGRAL FORMULATION 12
BOUNDARY ELEMENT FORMULATION 16
Discretization 16
Boundary Element Equations 18
Formation of System Matrices 19
Evaluation of Displacements 21
Summary '" , 21
DISCRETIZATION ERRORS 23
VERIFICATION 33
CONCLUSIONS . 41
REFERENCES 42
NOTATION 45
APPENDIX A: FREE-FIELD GROUND MOTIONS FOR SH, P, SV AND
RAYLEIGH WAVES 50
APPENDIX B: FULL-SPACE GREEN'S FUNCTIONS , 52
APPENDIX C: TRANSFORMED GREEN'S FUNCTIONS 53
Transformed Green's Displacement Functions 53
Transformed Green's Traction Functions 54
Expression for Matrix [C] 55
11l
PART II: IMPEDANCE FUNCTIONS FOR THREE-DIMENSIONAL
FOUNDATIONS SUPPORTED ON AN INFINITELY-LONG CANYON
OF UNIFORM CROSS-SECTION IN A HOMOGENEOUS HALF-SPACE
.............................. '" 57
INTRODUCTION 59
PROBLEM STATEMENT 60
FOURIER REPRESENTATION OF TRACTIONS AND DISPLACEMENTS 63
BOUNDARY INTEGRAL FORMULATION 65
BOUNDARY ELEMENT FORMULATION 68
Discretization 68
Boundary Element Equations 71
Formation of System Matrices 73
Evaluation of Displacements , 75
IMPEDANCE MATRIX 75
SUMMARY OF PROCEDURE 78
NUMERICAL ASPECTS 80
Selection of Sampling Points in Wavenumber Domain 80
Truncation and Discretization Errors 85
VERIFICATION AND COMPUTATIONAL COSTS 96
CONCLUSIONS 103
REFERENCES 105
NOTATION . 107
APPENDIX A: FOURIER TRANSFORM OF TWO-DIMENSIONAL SHAPE
FUNCTIONS 112
APPENDIX B: RELATION BETWEE~ {Uh AND {U}-k 116
APPENDIX C: RESTRICTIONS ON MESH LAYOUT ON f'c 119
IV
PART I
THREE-DIMENSIONAL ANALYSIS OF SPATIALLY VARY
ING GROUND MOTIONS AROUND A UNIFORM CANYON
IN A HOMOGENEOUS HALF-SPACE
INTRODUCTION
Spatial variations in earthquake motions around the canyon have rarely been con
sidered in practical earthquake analyses of arch dams, primarily for paucity of recorded
motions. Recent research studies have reported analyses of arch dam response to spa
tially varying ground motion determined from two-dimensional wave scattering analysis
[1-3]. Three-dimensional analyses would obviously lead to more realistic prediction
of how earthquake motions vary over the canyon. This report is therefore concerned
with such analyses to determine the spatial variations in earthquake motions around
canyons supporting arch dams.
The spatial variation of the ground motion resulting from the scattering and
diffraction of waves by a canyon has been the subject of many studies. Most of these
studies are concerned with the two-dimensional response of infinitely-long canyons
of uniform cross-section. The anti-plane shear response of such geometrically two
dimensional canyons to SH waves, parallel to the longitudinal axis of the canyon,
has been determined analytically for a semi-circular canyon [4], and a semi-elliptical
canyon [5]; and numerically for arbitrary cross-sections [6-8]. The numerical approaches
employed are based on the integral equation method [6], boundary method [7], and
finite element method [8].
The plane-strain response of geometrically two-dimensional canyons to P, SV and
Rayleigh waves normal to the canyon axis, which is much more complicated because of
mode conversions and coupled boundary conditions, does not seem to be amenable to
analytical solution. Consequently various numerical techniques have been developed.
Wong used a series of Lamb's solutions as trial function to represent the scattered field
and the complex amplitudes of the trial functions were determined through boundary
conditions in a least square sense by a generalized inverse method [9]. Sanchez-Sesma
et.al. applied Trefftz's wave function expansion method to construct the scattered
field [10]. The complex coefficients of the wave functions were then determined by
3
enforcing boundary conditions at both the canyon and the half-plane surface in a
least square sense. By expressing both the incident wave field and the scattered field
in terms of wave functions, Lee and Cao solved the diffraction of P and SV waves
at a two-dimensional circular canyon [11,12]. Their approach, however, results in a
set of infinite number of equations which can only be solved approximately. Various
boundary element methods (BEM), direct or indirect, have also been used to solve
the problem. Beskos et.al. analyzed the diffraction of waves by a trench using the
direct BEM with full-plane Green's functions [13]. An indirect BEM was applied
by Vogt et.al. to study wave scattering by arbitrary canyons in a layered half-plane
using half-space Green's function associated with distributed loads [14]. Nowak chose
the direct BEM with half-plane Green's functions associated with point loads [1]. By
combining the direct BEM with the discrete wavenumber Green's function, Kawase
investigated the time-domain response of a semi-circular canyon in a half-plane [15].
Hirose and Kitahara developed a special BEM to study the scattering of elastic
waves by inhomogeneous and anisotropic bodies in a half-space; the applicability of
their method, however, is limited because of the use of the fundamental solution of
elastostatics [16]. In addition to boundary methods mentioned above, various hybrid
methods (e.g. Mossessian and Dravinski [17]) have been used to study diffraction of
waves by surface irregularities.
Some of the numerical methods used for analysis of the plane-strain problem have
also been extended to study the diffraction of waves by finite-size, three-dimensional
surface irregularities including canyons and alluvial valleys. Wave function expansion
methods were applied by Sachez-Sesma et.al. [18,19], Eshraghi and Dravinski [20],
and Mossessian and Dravinski [21] to analyze the diffraction of waves by various finite
canyons and alluvial valleys. By expanding the incident and scattered fields in terms
of spherical wave functions, Lee analyzed the scattering of elastic plane waves by a
hemispherical canyon [22] and by a spherical cavity [23] in an elastic half-space. Using
the direct BEM, Banerjee et.al. studied wave barriers in multi-layered soil media [24]
4
and Rizzo et.al. investigated the scattering of waves by cavities in full-space [25,26].
The work on wave scattering and related problems has been reviewed by Wong [27],
Beskos [28] and Sanchez-Sesma [29].
Quasi two-dimensional formulations have been successful in analyzing three
dimensional responses for systems with cylindrical geometry [30-34J. Wong et. al.
studied the three-dimensional dynamic response of infinitely-long pipelines by using
a cylindrical eigenfunction expansion technique [30]. The ground motion due to a
steady state dislocation propagating along a fault of finite-width and infinite-length
were solved by Luco and Anderson for a homogeneous half-space [31J and by Mendez
and Luco for a layered half-space [32] through integral representation theory. Liu
et.al. investigated the three-dimensional scattering of seismic waves by a cylindrical
alluvial valley embedded in a layered half-space by using a hybrid method where
the boundary integral representation and FEM are combined in such a way that the
problem associated with the singularity of Green's functions is eliminated [33].
The three-dimensional response of an infinitely-long, uniform canyon subjected to
waves at an arbitrary angle with respect to the axis of the canyon has been analyzed
recently by Luco, Wong and de Barros [34]. This problem was solved for a canyon
cut in a layered viscoelastic half-space by the indirect boundary method utilizing the
Green's functions associated with a concentrated load moving along the canyon axis.
The number and cross-sectional location of these moving sources are determined by
trial and error and their magnitudes are computed by imposing boundary conditions
along the canyon surface in a least square sense.
In this report, we solve the same wave diffraction problem by a canyon, except
that the half-space is homogeneous instead of layered and Rayleigh wave excitation
is also considered, by the direct boundary element method in conjunction with full
space Green's function. The uniform cross-section of the infinitely-long canyon permits
closed-form integration along the canyon axis leading to a "two-dimensional" boundary
5
element problem involving Fourier transforms of the Green's functions. The assumption
of a homogeneous half-space seems to be appropriate for arch dam sites where similar
rock usually extends to considerable depth.
While we were working on this problem, an early, unpublished version of a paper
by Luco, Wong and de Barros [34J provided an independent set of results for evaluating
the accuracy of our results. In order to facilitate comparison, we follow their notation
and presentation in the next two sections and in the validation of numerical results.
SYSTEM CONSIDERED
The system considered consists of an infinitely long canyon of arbitrary but uniform
cross-section cut in a homogeneous viscoelastic half-space (Figure 1). The seismic
excitation is represented by P, SV, or SH waves incident on the canyon boundary
with arbitrary angles with respect to the vertical axis and the horizontal axes of the
canyon. Also considered are Rayleigh waves which approach the canyon horizontally
at an arbitrary angle with respect to the axis of the canyon. While appropriate
for exploratory purposes, free-field Rayleigh waves in a homogeneous half-space may
not be realistic since they imply larger amplitudes of the vertical component of the
wave than is supported by recorded data. Although the geometry of the canyon is
two-dimensional in that it is uniform along its length, the earthquake motion at the
canyon will vary along the axis of the canyon.
The viscoelastic half-space medium is characterized by the complex P and S wave
velocities Cp = cp(l + iTJp)I/2 and Cs = cs(1 + iTJs)I/2. The terms TJp and TJs represent the
constant hysteretic damping factors for P and S waves respectively; and cp and Cs
denote the P and S wave velocities if the medium were undamped. Similarly, the
Rayleigh wave velocity is given by Cr = cr(1 + iTJr )1/2 where Cr is the Rayleigh wave
velocity without damping and TJr is the corresponding hysteretic damping factor. Both
Cr and TJr are related to cp, cs, TJp and TJs.
6
Fig.! Infinitely-long canyon of arbitrary but uniform cross-section; rc is thecross-section of the canyon at x = 0; r i is the interface between thedam and foundation rock.
7
FREE FIELD MOTION
The first step is to determine the ground motion for free-field conditions, i.e. III
the half-space without the canyon. Considered first are the body waves. The seismic
excitation is represented by plane P, SV or SH waves in the underlying half-space
approaching from any direction. The normal to the wave front forms an angle 8v
with the vertical axis and an angle 8h with the axis of the canyon (Figure 2). The
incident motion in the underlying half-space, at any point xi = (x', y', z'), is described
by the three components of displacements given by
{U'(Xininc = A{u'}e-ik'x'+il/Z'+iwt (1)
where A is the amplitude of the incident displacement, w is the frequency of the
wave, k' = (wfcp)sin8v and v' = (wfcp)cos9v · for P-waves, and k' = (wfc s )sin8v and
v' = (wfcs)cos8v for S-waves. In equation (1), {u'} is the vector
for P-wavesfor SV-wavesfor SH-waves
(2)
The incident motion associated with Rayleigh surface waves, which propagate
hori- zontally, is given by
(3)
where A is a measure of the wave amplitude, k' = (wfc r ), and {u'(z')} is the vector
The free-field ground motion is the total motion in the half-space resulting from
the incident motion. The total motion due to surface waves is the same as the incident
motion (equation 3). In the case of body waves, the total motion is the incident
motion (equation 1) plus the reflected wave motion. The resulting total free-field
8
x
, x'~,, ,, ,, ,
, ', ', ", ,,~ffi, 7/oJ:
---F------:-,tr--~-_y, ,, ,, ,, ,,,,,,
z'__--~
r-Tir--'"""-----y------r-----~ X'
Rayleigh wave...
Fig.2 Canyon of uniform cross-section subjected to incident waves.
9
ground motion presented in Appendix A is well kn'own for the homogeneous half-space
[e.g. 35] but requires extensive calculation for a layered half-space [e.g. 36, 37]. In
either case, the free-field displacement field in the x'y'z'-coordinate system is denoted
by
(5)
where {u~fJ(z')} are the displacement components at the plane x' = O. In equation
(5), and in the sequel, the factor eiwt has been deleted for brevity. Similarly the
dependence of displacements and tractions on the frequency w is implied.
The free-field displacement is next transformed to the xyz-coordinate system in
which the boundary conditions at the surface of the canyon are imposed. For this
purpose, the rotation of coordinates is introduced:
(6)
The free-field displacements at any point x= (x, y, z) are
(7)
where {uJJ(io )} describes the displacements at i o = (0, y, z) on the y - z plane, k =(w/cp)sinOucos(h for P-waves, k = (w/cs)sinOucosOh for S-waves, and k = (w/cr)COSOh
for Rayleigh waves. Equation (7) suggests that waves travelling in the x'-direction
with speed c described by equation (5), may be interpreted as propagating in the
x-direction with an apparent speed of C/COSOh'
In the presence of the canyon, the total displacement field {u( i)} consists of the
free-field motion {uJJ(i)} and the scattered-field motion {us(i)}:
(8)
Similarly, the total traction vector {t (i)} on the canyon boundary and the half-space
surface is
(9)
10
where {tjj(xn and {ta(xn are the traction vectors associated with the free-field and
scattered field respectively, and n denotes the normal to the canyon and half-space
boundary at x.
The traction vector {tfJ(xn can be determined from the displacement field {ufJ(x)}
through the strain-displacement relations and material constitutive law. It can be
expressed as
(10)
The traction free condition on the boundary r c of the canyon requires that (using
equation 9)
(11)
On the boundary r h of the half-space, the same condition leads to
(12)
because the free-field solution satisfies the traction free condition.
To determine the total displacements, {u(x)}, around the boundary of the canyon,
only the scattered field in equation (8) needs to be evaluated since the free-field
displacements are known from equation (7). The scattered displacement field is of
the form
(13)
Thus, once the displacements along the boundary rein the plane x = 0 are determined,
their values at the dam-foundation rock interface are known from equation (13).
Thus, the problem reduces to evaluating the displacements along the canyon
boundary at the x = 0 plane associated with tractions at the canyon boundary and
half-space surface that are known from equations (11) and (12). This problem is solved
in the subsequent sections by the boundary element method, modified to recognize
the uniform cross-section and infinite length of the canyon.
11
BOUNDARY INTEGRAL FORMULATION
The boundary integral equation can be obtained from the reciprocal theorem
applied to a pair of elastodynamic equilibrium states; this theorem is a dynamic
extension of the classical Betti's law for elastostatic systems. Both the selected elas
todynamic equilibrium states satisfy the wave equation over the half-space with a
cut canyon and radiation conditions at infinity, but not necessarily the boundary
conditions. The first state is the unknown scattered displacement field {us(x)} and
the associated tractions {t(x)} at the half-space surface and canyon boundary, which
can be determined from equations (11) and (12). The second state is defined by the
displacements {u*(x,xos)} and tractions (t*(x,xos)} in a full-space due to concentrated
forces {P} = {Px , Py, Pz}T applied at a point xos = (0, Ys, zs) on f' c or f'h, the intersection
of the system boundary and the x = 0 plane. According to the reciprocal theorem,
the work done by the tractions of the second state in undergoing the displacements
of the first state is equal to the work done by the tractions of the first state through
the displacements of the second state:
The integration is over the entire boundary S consisting of the surface of the half
space and the boundary of the canyon, modified slightly to avoid the singularity at
the point of load application (Figure 3).
nThe displacements {u*(x, xos)} and tractions {t*(x, xos)} at a "receiver" point x =
(x, y, z) on the boundary S are related to the concentrated forces applied at the
"source" point xos = (0, Ys, zs) on the boundary f' c and f\ :
(lSa)
(lSb)
The traction vector is associated with the normal n to the boundary, and [G*(x, xos)]n
and [P'(X, xos)] are 3 x 3 matrices of Green's functions. The first, second and third
12
Fig.3 Definition of boundaries and source point; r c is the cross-section ofthe canyon at x = 0, rh is the x = 0 line on the surface of the halfspace, and r. is the small contour of radius r. around the sourcepoint.
13
ncolumns of the matrices [G*(x, xos)] and [F*(i, xos)] 'correspond to the displacement and
traction vectors at x for a unit point load acting at xos in the x, y and z directions,
respectively. The 3 x 1 vector {P} represents the unknown amplitudes of the forces.n
Expressions for matrices [G*(x, xos)] and [F*(x, xos)] are presented in Appendix B [38].
The boundary integral equation is obtained by substituting equation (15) into
equation (14) and recognizing that the latter should be satisfied for all force vectors
{P}=
[ /00 [F*(x,xos)f{us(i)}dxdf =Jrcur"ur. -00~ _/00 [G*(i, ios)f{t(i)}dxdf
Jrcur"ur. -00 (16)
For each source point xos, this equation represents three scalar equations in the
unknown displacements {us(i)}.
Equation (16) can be simplified by substituting equations (10), (11) and (13) and
analytically evaluating the integral along the x-axis, leading to
(17)
(18b)
(18a)
where
[G(io, xos)] = 1:[G*(x, xos)]e-ikXdx
[F(xo, Xos)] = 1:[F*(x, xos)]e-ikxdx
Expressions for these Fourier transforms are given in Appendix C. Thus the original
boundary integral equation (16) has been transformed to a simpler form (equation
17) in which the remaining integration is only along the curve l'c u l'h U 1's at the
x =0 cross-section of the system.
The final step is to make the radius rs of the small arc 1's around the source
point, ios, approach zero. Following [39], it can be shown that
(19)
14
(21)
where
[C( ios)] = lim [[F( i o, ios)fdf (20)r.-olr.
A general expreSSIOn for the 3 x 3 matrix [C(ios)), which depends on the geometry
of the boundary at the source point, is presented in Appendix C. If the boundary is
smooth, i.e., it has a unique tangent, at the source point, Cij(ios) =Cij/2.
The final form of the boundary integral equation is obtained from equation (17)
after utilizing equation (19):
[C(ios)]{us(ios)} + [ [F(io, ios)]T{us(io)}dflreur"
= - ~ _ [G(io, xos)]T{fff(io)}dflrc ur"
nIn this equation, the traction vector {t ff (i o)} is known from equation (10) and the
ntransformed Green's function matrices [G(io,ios)] and [F(io,xos)] are known from
equation (18) and Appendix C, whereas the scattered-field displacements at the plane
x = 0 are the unknowns. These include {us(ios)} at the source point and {us(xo )} at
the rest of the boundary feu l'h. Once the displacements along the boundary feU f h
are determined, the displacements at the rest of the boundary f, and in particular
at the dam-foundation rock interface can be determined from equation (13).
By taking advantage of the uniform geometry of the system along the x-direction,
the three-dimensional boundary integral equation (16) has been reduced to a "two
dimensional" problem (equation 21) involving Fourier transforms of Green's functions.
This was possible because the x-variation of tractions and displacements is known
analytically (equations 11 and 13) which permitted closed-form integration along the
x-axis. The reduced problem is really not two-dimensional because the unknown
scattered displacements along the boundary feU f h are three-dimensional vectors.
If the incident waves are perpendicular to the x-axis, Oh = 900 and therefore k =0,
equation (18) becomes
(22a)
15
(22b)
nIn the 3x3 matrices [Go(xo,xos)] and [Fo(xo,xos)], the (1,1) entry represents the Green's
functions for the 2-D anti-plane shear motion; the (2,2), (2,3), (3,2) and (3,3) entries are
the Green's functions associated with the 2-D plane-strain motion, and the remaining
entries are zero because the in-plane and anti-plane motions are uncoupled. In this
case, the boundary integral equation governs two uncoupled problems: the anti-plane
shear problem associated with an incident SH wave and the plane-strain problem
associated with incident P, SV and Rayleigh waves.
BOUNDARY ELEMENT FORMULATION
Discretization
In its present form, equation (21), which represents the exact formulation of
the problem, cannot be solved analytically to determine the scattered displacements,
even for the simplest canyon geometry. Therefore, the integration domain f'c u f' h is
discretized into M one-dimensional elements (Figure 4). Obviously, the discretization
extends to only a finite distance along f' h on the half-space surface, which should be
large enough to provide accurate solutions. In the simplest form, each element has
two nodes, and the interpolation functions are linear. Following isoparametric element
concepts, these interpolation functions are Nt (() = 0.5(1 - () and N 2(() = 0.5(1 +()where ( varies from -1 to 1 (Figure 4). Thus the discretized system with M elements
has M n = M +1 nodes.
The variation of the scattered-field displacements over the jth element can be
expressed in terms of nodal displacements
where [N(()] is a 3 x 6 matrix consisting of interpolation functions
(23)
(
Nt[N(()] = ~ (24)
16
FhelementJ
o
z
y
Fig.4 Boundary element discretization.
17
and {iT}; is the vector of displacements at the two nodes - j and j + 1 - of the
element
Boundary Element Equations
The boundary integrations in equation (21) can now be expressed as summation
of integrals over all the M elements
M
[C(xos)]{us(xos)} +2:=i [F(Xo,Xos)]T{us(xo)};df;=1 t j
(25)
where 1"; represents the extent of the jth element. The summation on the right side
extends over only the elements of fe, the canyon boundary (Figure 4), because the free
field tractions {tff(xo )} vanish on f h , the half-space surface, and hence the contribution
of the corresponding elements is zero. By expressing the displacement field in terms
of nodal displacements through equation (23), and changing the integration variable
to (, the contributions of the jth element in equation (25) become
[ [F(xo,xos)];{us(xo)};df = ~Ijjl [F«(,xos)]J[N«()]d({iT}j (26a)itj -1
[ [G(xo, xos)];{tfJ(xo)}jdf = -211j /1 [G«(, xos)];{tfJ«()}jd( (26b)~j -1
where I; is the length of the jth element. Substituting these equations, equation (25)
can be expressed in matrix form:
Equation (27) represents three equations associated with the source point xos
(0, Ys, zs), where the 3 X 3 matrix [C(xos)] is given by equation (20), the 3 x 1 column
vector {us(xos)} represents the unknown displacements at the source point, {iT} is a
18
3Mn x 1 column vector of unknown nodal displacements (3 displacement components
at each of the M n nodes), [A(xos)] is a 3 x 3Mn matrix assembled from the integrals of
the products of the transformed traction Green's functions and element interpolation
functions:- 1
[A(xos)] = 2M 1
~ I j 11 (F«(,xos)]J[N«()]d(a.s.semble
(28)
and {Q(ios)} is a 3 x 1 vector defined by
(29)
The matrix [A(xos)] is obtained by appropriately assembling the contribution of the
various elements, a process that is reminiscent of the stiffness matrix assembly process
in the direct stiffness method. The vector {Q(xos)} is obtained by direct addition of
the contributions of the various elements. The construction of both of these matrices
for a small system is illustrated in Figure 5.
Equation (27) is a discrete form of equation (21) obtained by introducing two
approximations: (1) interpolation of displacements over each element in terms of nodal
displacements (in this particular formulation the interpolation functions were chosen
to be linear); and (2) truncation of the integration over f h to a finite distance.
Formation of System Matrices
Associated with the source point xos = (0, Ys, zs), equation (27) represents 3 linear
algebraic equations in 3Mn unknown displacements. Obviously a total of 3Mn equations
are needed to determine the unknown displacements. These equations can be obtained
by successively choosing the source point at each of the nodes of the discretized
boundary. When the source point coincides with the ith node, Xi, equation (27)
becomes
\.
(30)
Such equations associated with all nodes can be obtained from equation (30) by
19
•5
++++ooofoofoofoo~ ••~ ••
ooofoofoofoo~··r··000000000000 ••••••
•
}{U}
•~+~+t:+~+~
contribution ofelement 3 = {Q(xos)h
...~..~..~..~..~..···~··~··I···~··~··••••••••••••••••••
Fig.5 Assembly of Matrices.
20
varying i = 1,2, ... , M n and assembling the resulting equations in the following form:
[AHU} = {Q}
where [A] is a square matrix of order 3Mn given by
(31)
(
[C(id]
[A] = (32)
(33)
and {Q} is a 3Mn x 1 column vector
{Q} = ( {Q(Xl)) 1{Q(iMn )}
The coefficient matrix [A] is unsymmetric and fully populated, which is typical of the
boundary element method equations.
Evaluation of Displacements
Solution of the system of linear algebraic equations (31) leads to the nodal dis
placements {U} and thereafter the displacement variation over each finite element from
equation (23). The resulting displacements represent an approximation to {u,,(io )},
the scattered displacement field along the boundary f'c u f' h at the plane x = 0. The
scattered displacements {u,,(in on the boundary surface at any other plane are then
given by equation (13), which when combined with the known free-field displacements
(equation 7) lead to the total displacements, {u(x)} (equation 8).
Summary
The boundary element procedure developed m the preceding sections may be
summarized as follows:
1. Discretize f'c, the canyon boundary at x = 0, and f'h, the half-space surface at
x = 0, into M line elements. If linear interpolation functions are selected for each
element, the discretized system contains Mn = M +1 nodes.
21
2. Establish equation (30) which includes three linear algebraic equations associated
with the ith node, X;, in 3Mn unknowns (3 displacement components at each of
the M n nodes) by the following steps:
(a) Compute the elements of the 3x3 matrix [C(x;)] which depend on the geometry
of the boundary at the ith node, see Appendix C. These elements are given
by Cjk(xd =Ojk/2 if the boundary is smooth, i.e., it has a unique tangent at
X;.
(b) Compute the 3x3Mn matrix [..4(i;)] from equation (28) by assembling thecontri
butions of all the M elements, as described in Figure 5. Element contributions
are determined from the integrals of the transformed traction Green's functions
(equation I8b) and element interpolation functions. Analytical expressions for
the transformed traction Green's functions are available in Appendix C.
(c) Compute the 3 x 1 vector {Q(id} from equation (29) by adding the contri
butions of the elements on the canyon boundary, as described in Figure 5.
Element contributions are determined from the integrals of the products of
transformed displacement Green's functions (equation 18a) and free-field trac
tions. Analytical expressions for the transformed Green's functions are avail
able in Appendix C and the traction vector is determined from the free-field
displacements {uff (i)} given by equation (7) through the strain-displacement
relations and material constitutive law.
3. Repeat the computations summarized in step 2 for each of the nodes, i = 1,2, ... , M n
to determine [C(i;)], [..4(i;)] and {Q(i;)} for all the nodes.
4. Evaluate the square matrix [A] of order 3Mn and vector {Q} of size 3Mn for the
boundary element system from the corresponding individual nodal matrices (step
3) using equations (32) and (33).
5. Solve the system of 3Mn linear algebraic equations to determine the nodal dis
placements {U} and thereafter the displacement variations over each finite element
22
from equation (23).
6. Determine the scattered displacements {u s ( x)} at the dam-foundation rock inter
face, or any plane other than x = 0, from the displacements at the x = 0 plane
(step 5) using equation (13).
7. Add the scattered displacements {us(x)} from step 6 and the free-field displace
ments {uff(x)} , determined from equation (7), to obtain the total displacements
{u(x)}.
A computer program has been developed to implement the procedures summarized
above for an incident wave of frequency w with its propagation direction defined by
angles (Jv to the vertical, z-axis and (Jh to the horizontal, x-axis of the canyon.
The boundary element formulation summarized above is for a canyon cut m a
homogeneous viscoelastic half-space. In principle, this method can be extended to a
layered half-space with non-horizontal layers. A boundary integral equation, similar
to equation (21), can be obtained for each layer and the underlying half-space. The
interfaces between layers would also require discretization because full-space Green's
functions are used. Thus this approach may become computationally impractical for
multi-layered systems.
DISCRETIZATION ERRORS
As mentioned earlier, discretization of the boundary obviously extends only a finite
distance, Lh' along the half-space surface at the plane x = 0 (Figure 4). The minimum
Lh necessary for accurate solution of the problem is examined next. Computed by
the present procedure, with different L h values, the displacement amplitudes around
a semi-circular canyon of radius L are presented in Figures 6 to 9 for P, SH, SV
and Rayleigh waves for incidence angles fh = 30° and (Jv = 45° and three values of
the normalized frequency, n =wL/'lrcs • For L =800 ft, and Cs =6000 ft/sec, n =0.5,1
and 2 represent frequencies f = 1.9,3.8 and 7.5 Hz. The results are presented for four
selections of Lh : L, 2L, 4L and 6L; in each case the element size is taken as L/8. The
23
4. = 1L4. = 2L
- - - 4. = 4L4. = 6L
3 --.--------..,o = 0.5..
.2 2Q)"0~
~Q.E 1<.Q.m
C 0 -+---r---r-"""T""""""T"""....-....--.---i
3 .........---------,o = 0.5
o = 1.0 o = 2.0
o = 2.0
o = 0.5 o = 2.0o = 1.0
.Q.m
C 0 -+---'---'-"""T"""--r---r---;---;--;
3--.---------.,
.""
.2 2Q)"0~
.~Q.E 1<
..
..:! 2Q)"0~
~Q.E 1<
-1 0 1Distance y/L
2 -2 -1 0 1Distance y/L
2 -2 -1 a 1Distance y/L
2
Fig.6 Influence of Lh on computed response of a semi-circular canyon toP-wave excitation (fh = 30°, ()v = 45°).
24
4t = 1LL,. = 2L
- - - L,. = 4LL,. = 6L
2
{] = 2.0
{] = 2.0
-1 0 1Distance y/L
2 -2
{] = 1.0
{] = 1.0
{] = 1.0
-1 0 1Distance y/L
2 -2
0.5
-1 0 1Distance y/L
(] = 0.5
{] = 0.5
3 ........----------,
Q.encO-+---.---.---.---,.--,.~___.--I
3 ........----------,
-;..::J- 2enQ)1J::J
:e=.
E"1<
.CiencO-+--.---r--r----,----r--...,.---r~
-2
CiencO-+---.---.---,.~___.___,,...-r_l
3 -,--------...,
..::J- 2enQ)1J::J~
E"1<
..::J- 2enQ)
1J::J~
E"1<
Fig.7 Influence of Lh on computed response of a semi-circular canyon toSH-wave excitation (fh = 30°, Bv = 45°).
25
Fig.8 Influence of Lh on computed response of a semi-circular canyon toSV-wave excitation (lh = 30°, Bv = 45°).
26
------------ Lh = 1L - - - Lh = 4L- - -- ~ = 2L ~ = 6L
30 = 0.5 0 = 1.0 0 = 2.0..
:l2
enII)"0:l~
E1<.0-enC 0
30 = 0.5
-...:l
2enQ)
"0:l~
Q.E<Q.enC 0
40 = 0.5 0 = 1.0 0 = 2.0
..;3 /h\en / ", ,
II)I' ,
"0 2 II ,,~,
:l I
-.J
'~" AQ.E ~., , ,< .......
Q.en
CS 0-2 -1 0 1 2 -2 -1 0 1 2 -2 -1 0 1 2
Distance y/L Distance y/L Distance y/L
Fig.9 Influence of L h on computed response of a semi-circular canyon toRayleigh wave excitation (fh = 30°).
27
semi-circular curve fe, which is the intersection' of the canyon and plane x = 0, IS
discretized into 24 elements.
It is apparent from the results of Figures 6 to 9 that, in most cases, the displace
ments on the canyon surface are not very sensitive to the Lh value. The dependence
on Lh decreases for higher frequency. It is more meaningful to define the required Lh
as a multiple of the shear wavelength. Such results demonstrate that if Lh exceeds
twice the shear wavelength the results are accurate.
The dependence of the boundary element solution on the fineness of the dis
cretization is examined next. Computed by the present procedure, the displacement
amplitudes around a semi-circular canyon are presented in Figures 10 to 13 for P,
SH, SV and Rayleigh waves for normalized frequency n = 0.5,1,2 and (Jh = (Jv = 45°.
The corresponding shear wave lengths are equal to 4L, 2L and L where L is the
radius of the canyon. The canyon boundary is divided into N e elements and the
half-space surface into N h elements with the latter covering an extent of L h = 4L on
both sides of the canyon. The element size on f h is about the same as that on f e •
The results obtained from three different discretizations are shown in Figures 10 to
13 together with what might be considered as the "exact" result. The latter was
obtained with very fine discretization, N e = 32 and N h = 40. It is apparent that, as
the discretization becomes finer, the results converge to the "exact" values. A coarse
mesh gives reasonably accurate results only at low frequency, and a finer mesh must
be used at high frequency to describe the rapid variations of the displacements. It
is the ratio of the shear wavelength to the element size that determines if a chosen
mesh is fine enough. Reliable results are obtained when this ratio is greater than 5.
In summary, the present method is reliable and can be used to analyze the spatial
variations of motion around an infinitely-long canyon of arbitrary shape. In order to
control discretization errors, the element size on the canyon surface should be kept
smaller than one-fifth of the shear wavelength, and the half-space surface should be
28
o Nc=4, Nh=5Nc=8, Nh =10
Nc= 16. Nh=20-- Nc=32. Nh=40
3--.------------,o = 0.5..
..2 2CD'0:J
;'=Q.E 1<Q.en
° O-+----r---r---r---r---r---.-.---l
3---r----------,o = 0.5
o = 1.0
o
o = 1.0
o = 2.0o
o
o = 2.0
o = 1.0o = 0.5
-,..:J 2CD'0:J
;'=Q.E 1<Q.en
°O-+---.-r---r--..,...--r---r----T_
3--.------------,
..:J 2CD'0:J
.":a.E 1<
2-1 0 1Distance y/L
2 -2-1 0 1Distance y/L
2 -2-1 0 1Distance y/L
Q.en
°O-+---r----r---r-....---,r--r--,.........-l
-2
Fig.l0 Influence of mesh refinement on computed response of a semi-circularcanyon to P-wave excitation ((h = (Jv = 45°, Lh = 4L).
29
o Nc=4, N..=5Nc=8, N..=10
Nc= 16, N..=20-- Nc=32, N,,=40
4--.---------,o = 0.5 o = 1.0 o = 2.0
·3:Jo
o
CD"0:J 2
;0=Q.E< 1.Q.(I)
is o-+-,--,--,-...,.--r-...-"""'"4--.----------.,
o = 0.5 o = 1.0 o = 2.0
o = 0.5
2
o = 2.0
o
-1 0 1Distance y/L
2 -2
o
o = 1.0
-1 0 1Distance y/L
2 -2-1 0 1Distance y/L
- .. 3..2CD"0:J 2
;0=Q.E<1.Q.(I)
is 0-+-,--,--,-...,..........,.....................-13--r---------,
Q.(I)
iSO-+-.----,.l~__,.___r--,___,,..._'
-2
..:J 2CD"0:J
;0=Q.E 1<
Fig.ll Influence of mesh refinement on computed response of a semi-circularcanyon to SH-wave excitation (Oh = Ov = 45°, Lh = 4L).
30
o Nc=4, Nh=5Nc=8, Nh=10
- - Nc=16. Nh=20-- Nc=32. Nh=40
3--r----------,a = 0.5 a = 1.0 a = 2.0
2
o
o
o
a = 2.0
a = 2.0
o
o
o
2 -2 -1 0 1Distance y/L
a = 1.0
2 -2 -1 a 1Distance y/L
..:)
2GJ
'U 0:);0=Q.E 1<a.(I)
i5 a3
a = 0.5
-,..,:)
2GJ'U:)~.-Q.E 1<a.(I)
i5 a4
a = 0.5
.. 3:)
GJ'U:) 2~
Q.E<a.(I)
c a-2 -1 a 1
Distance y/L
Fig.12 Influence of mesh refinement on computed response of a semi-circularcanyon to SV-wave excitation (fh = f)v = 45°, Lh = 4L).
31
Fig.13 Influence of mesh refinement on computed response of a semi-circularcanyon to Rayleigh wave excitation (fh = 45°, .Lh = 4L).
32
discretized on both sides of the canyon over a distance of at least two times the shear
wavelength.
VERIFICATION
The analytical procedure and the implementing computer program were tested by
solving problems for which previous results are available. In all cases, an infinitely
long canyon with semi-circular cross-section of radius L cut in a homogeneous half
space is considered. The earlier results [4,9] for the two-dimensional response of the
canyon are for a elastic half-space with cp = 2cs , i.e. Poisson's ratio = 1/3. The
recent results [34] for the three-dimensional response of the canyon are for a slightly
dissipative viscoelastic half-space characterized by cp = 2cs and TIp = TIs = 0.02. The
present approach is implemented for the latter set of material properties, with the
canyon boundary f'c discretized into 32 equal elements and each side of the half-space
boundary f'p is discretized into 30 equal elements covering Lh =4L (Figure 4).
The results obtained by the present method for the two-dimensional anti-plane
shear motion arising from an SH wave of unit amplitude impinging normal to the
canyon axis are compared in Figure 14a with the exact solution of Trifunac [4]. The
results presented describe the displacement amplitudes around the canyon for Ov =
1°,45° and 85° and a dimensionless frequency n = wL/1rcs = 1. A similar comparison
is presented in Figure 14b for Ov = 0° and Oh = 0° which corresponds to a vertically
incident SH wave with particle motion perpendicular to the axis of the canyon. For
this plane-strain case, the numerical results are available (e.g. Wong [9]). Figure 15
shows the results for the two-dimensional response of the canyon to P and SV waves
impinging normal to the axis of the canyon (Oh = 90°) at various vertical angles Ov
and for a dimensionless frequency n = wL/1rcs = 1. The comparisons in Figures 14
and 15 indicate that the present method provides reasonably accurate results for the
two-dimensional limiting cases considered.
We now turn to the three-dimensional response of a canyon to incident waves
33
4--,-----------------------~
3
-.::l
V"U::l 2~
0.E«
0.1enC5
?9
r:I/
o b jJ'0
17., = 1°- - - - 17., = 45°--------- 17., = 85°
1.00.50.0Distance y/L
-0.5O-+----r----r---,-,...--.--....,---,-~____r__,_-r__;____r__,__.___r_,__,-r_i
-1.0
4 """""T"""--------------------------,----Iurl- - - - IUrl
3
2oDistance y/L
-1
II
90'D-tJ g,.O.9
O-+---..-.....--..,..--...,.....--r---,.-.,..---.--4It--.-...---.--..---,.-,.......-.....--r----r---1
-2
0.en
C5
Fig.14 Comparison of results of present method (lines) with previousresults (Ref. [4] shown by 0, and Ref. [9] shown by c ) for SHwave excitation; n = 1.
34
P-waves SV-WQves
3-,---------------. 4---r------------~
,'O°'Qoa> ,I \I \
3.....,....---------------.~ =300
I .,....I
I
i 2JJ1Ci
E 0« 1\'Ci1~
i5 J
4-,-------------~
3
2
3-,-------------, 10 """T""-------------..
2-1 0Distance y/l
2
6
4
8
O+-,.,........_r...,....,~T"'""T""...._r...,....,~l,__r_"'T""'T_,_\
2 -2I
-1 0Distance y/l
o I
-2
Fig.IS Comparison of results or present method (lines) with previousresults (Ref.[9J shown by 0) for P- and SV- wave:! excitations;n = 1, ()h = 900
•
35
impinging from an arbitrary direction. The calculated amplitudes of the three dis
placement components (lu,,;I,luyl, and luzl) for incident P, SH and SV waves of unit
amplitude, dimensionless frequency n =wL/1rcs =1, vertical angle of incidence Ov =45°,
and horizontal angle of incidence Oh = 45° are presented as solid lines in Figure 16
for an infinitely-long canyon of semi-circular cross-section, dashed lines represent the
results obtained by the indirect boundary method [34]. This comparison indicates that
the present method provides reasonably accurate results also for the three-dimensional
problem.
Also included In Figure 16 are the displacement amplitudes around the canyon
computed by analyzing a finite-length canyon by a fully three-dimensional verSIOn
of the present boundary element method. In this case, discretization is necessary
even along the canyon axis (Figure 17) in contrast to the method presented in the
preceding sections where integrations in the x-direction could be evaluated analytically
(equation 14). The results at x = 0 obtained for the finite-length canyon of Figure 17
with d =4L and Lh = 1.5L are shown in Figure 16 by discrete symbols. The finiteness
of the canyon has only a small effect on the displacements far from the ends of the
canyon. The results for the finite-length canyon can be improved in accuracy at the
expense of increased computational effort - by increasing Lh and by using a finer
discretization on the canyon boundary.
The results presented in Figure 18 are for an infinitely-long canyon (Figure 1) and
finite canyons (Figure 19) of four different lengths subjected to Rayleigh waves with
dimensionless frequency n = wL/1rcs = 1, and horizontal angles of incidence 0h = 30°
and 60°. This comparison indicates that the results obtained for infinite and finite
canyons are generally similar and become closer with increasing length of the finite
canyon. The accuracy of the results obtained for the finite-length canyon is limited
because of the coarseness of the discretization on the canyon boundary and the short
distance, Lh = 1.5L, over which the discretization extends on the half-space surface
(Figure 19). Compared with the fully three-dimensional version of the boundary
36
2
SV-woves
-1 0 1Distance y/L
2 -2
o
SH-waves
o
2 -2 -1 0 1Distance y/L
-2
P-waves
2
2
o -+-----,I,---,-~_,__,_-.-.....,
-1 0 1Distance y/L
O--+--r---r---r---r---r---r---r--l
3--r-----------,
2
3-,------------,
O-+-~_r__,____r_,___r__,--1
3-.------------.,
Fig.16 Comparison of results of present method (continuous lines) with previous results (Ref.[34] shown by dashed lines) for P-, SH- and SV-waveexcitations (n = 1.0, (h = ()v = 45°, "7P = "7s = 0.02). Results forfinite-length canyon (d = 4L) are shown by discrete symbols.
37
z
x
x
l
d
l
Fig.17 Canyon of finite-length.
38
y
2
3~-------------.,
•;lF.....I
o !
0 t 0
3 J 317h=30o "h=60;
IJ
i2-1 2
IUyl I1 •
c• 0c0
2-1 aDistance y/L
O-+...---.~...___."""'T_"...___.__r_T""""T_,_T""""T_,_,._,.....-rl
-2
2
4 -T'"----------------,
3
2
•8
8
-1 0Distance y/L
o i
-2
4.....,....--------------,J 19h=30°
I
3..JIi
....i
Fig.IS Comparison of results for infinite canyon by the present method (lines)and for finite canyons of different lengths (d = L shown by 0, d = 2Lby c , d = 3L by [> and d = 4L by *) by the 3-D BEM for Rayleighwave excitation (0 = 1, 77p = 77. = 0.02).
39
x
..J
..Jas@)
ECDECD
Q)
EJcII
"'C
L 1.5L
n m a1 4 0.2502 6 0.3333 8 0.3754 8 0.500
y
Fig.19 Top view of a quarter of the discretized system for finite canyons offour different lengths.
40
element method, the present approach is much more efficient as it permits analytical
integration along the canyon axis.
CONCLUSIONS
A direct boundary element procedure to calculate the three-dimensional response of
an infinitely-long canyon of arbitrary but uniform cross-section cut in a homogeneous
viscoelastic half-space to incident seismic waves has been presented. The seismic
excitation is represented by P, SH, SV or Rayleigh waves at arbitrary angles with
respect to the axis of the canyon. The uniform cross-section of the infinitely-long
canyon permits analytical integration along the canyon axis of the three-dimensional
boundary integral equation. Thus the original three-dimensional problem is reduced
to a computationally efficient "two-dimensional" boundary element problem involving
Fourier transforms of full-space Green's functions.
The accuracy of the procedure and implementing computer program has been ver
ified by comparison with previous solutions for the limiting case of two-dimensional re
sponse, recently obtained three-dimensional response results for infinitely-long canyons
by the indirect boundary method, and three-dimensional boundary element method
solutions presented here for finite canyons. The finiteness of the canyon has only a
small effect on the displacements far from the ends of the canyon. Compared with the
fully three-dimensional version of the boundary element method, the present method
is much more efficient as it permits analytical integration along the canyon axis. This
method is suitable for research studies concerning the influence of spatial variations
in input motions on the earthquake response of arch dams.
41
REFERENCES .
1. P.S. Nowak, "Effect of nonuniform seismic input on arch darns," Report No. EERL88-03, California Institute of Technology, Pasadena, California, 1988.
2. S.B. Kojic and M.D. Trifunac, "Earthquake stresses in arch darns. I: theory andantiplane excitation," J. Engineering Mechanics, ASCE, 117, No.3, pp.532-552,1991.
3. S.B. Kojic and M.D. Trifunac, "Earthquake stresses in arch darns. II: excitationby SV-, P-, and Rayleigh waves," J. Engineering Mechanics, ASCE, 117, No.3,pp.553-574, 1991.
4. M.D. Trifunac, "Scattering of plane SH waves by a semi-cylindrical canyon,"Earthquake Engineering and Structural Dynamics, 1, pp.267-281, 1973.
5. H.L. Wong and M.D. Trifunac, "Scattering of plane SH wave by a semi-ellipticalcanyon," Earthquake Engineering and Structural Dynamics, 3, pp.157-169, 1974.
6. H.L. Wong and P.C. Jennings, "Effects of canyon topography on strong groundmotion," Bull. Seism. Soc. Am., 65, pp.1239-1257, 1975.
7. F.J. Sanchez-Sesma and E.Rosenblueth, "Ground motion at canyons of arbitraryshape under incident SH waves," Earthquake Engineering and Structural Dynamics,7, pp.441-450, 1979.
8. W.D. Smith, "The application of finite element analysis to body wave propagationproblems," J. Geophys., 44, pp.747-768, 1975.
9. H.L. Wong, "Effect of surface topography on the diffraction of P, SV, and Rayleighwaves," Bull. Seism. Soc. Am., 72, pp.1167-1183, 1982.
10. F.J. Sanchez-Sesma, M.A. Bravo and 1. Herrera, "Surface motion of topographicalirregularities for incident P, SV, and Rayleigh waves," Bull. Seism. Soc. Am., 75,pp.263-269, 1985.
11. H. Cao and V.W. Lee, "Scattering and diffraction of plane P waves by circular cylindrical canyons with variable depth-to-width ratio," Soil Dynamics andEarthquake Engineering, 9, No.3, pp.141-150, 1990.
12. V.W. Lee and H. Cao, "Diffraction of SV waves by circular canyons of variousdepths," J. Engineering Mechanics, ASCE, 115, No.9, pp.2035-2056, 1989.
13. D.E. Beskos, B. Dasgupta and 1.G. Vardoulakis, "Vibration isolation using openor filled trenches," Computational Mechanics, 1, pp.43-63, 1986.
14. R.F. Vogt, J.P. Wolf and H. Bachmann, "Wave scattering by a canyon of arbitraryshape in a layered half-space," Earthquake Engineering and Structural Dynamics, 16,pp.803-812, 1988.
15. H. Kawase, "Time-domain response of a semi-circular canyon for incident SV,P, and Rayleigh waves calculated by the discrete wavenumber boundary elementmethod," Bull. Seism. Soc. Am., 78, pp.1415-1437, 1988.
16. S. Hirose and M. Kitahara, "Elastic wave scattering by inhomogeneous andanisotropic bodies in a half space," Earthquake Engineering and Structural Dynamics,18, pp.285-297, 1989.
17. T.K. Mossessian and M. Dravinski, "Application of a hybrid method for scatteringof P, SV, and Rayleigh waves by near-surface irregularities," Bull. Seism. Soc.Am., 77, pp.1784-1803, 1987.
18. F.J. Sanchez-Sesma, "Diffraction of elastic waves by three-dimensional surfaceirregularities," Bull. Seism. Soc. Am., 73, pp.1621-1636, 1983.
42
19. F.J. Sanchez-Sesma, L.E. Perez-Rocha, and S. Chavez-Perez, "Diffraction of elasticwaves by three-dimensional surface irregularities. Part II," Bull. Seism. Soc. Am.,79, pp.101-112, 1989.
20. H. Eshraghi and M. Dravinski, "Scattering of Plane Harmonic SH, SV, P, andRayleigh waves by non-axisymmetric three-dimensional canyons: a wave functionexpanSlOn approach," Earthquake Engineering and Structure Dynamics, 18, pp.983998, 1989.
21. T. Mossessian and M. Dravinski, "Amplification of elastic waves by a threedimensional valley, Part I: Steady state response," Earthquake Engineering andStructural Dynamics, 19, pp.667-680, 1990.
22. V.W. Lee, "A note on the scattering of elastic plane waves by a hemisphericalcanyon," Soil Dynamics and Earthquake Engineering, 1, No.3, pp.122-129, 1982.
23. V.W. Lee, "Three-dimensional diffraction of elastic waves by a spherical cavityin an elastic half-space, I: Closed-form solutions," Soil Dynamics and EarthquakeEngineering, 7, No.3, pp.149-161, 1988.
24. P.K. Banerjee, S. Ahmad and K. Chen, "Advanced aRplication of BEM to wavebarriers in multi-layered three-dimensional soil media, Earthquake Engineering andStructural Dynamics, 16, pp.1041-1060, 1988.
25. F.J. Rizzo, D.J. Shippy and M. Rezayat, "A boundary integral equation methodfor radiation and scattering of elastlc waves in three dimensions," Int. J. forNumerical Methods in Engineering, 21, pp.115-129, 1985.
26. M. Rezayat, D.J. Shippy and F.J. Rizzo, "On time-harmonic elastic-wave analysisby the boundary element method for moderate to high frequencies," ComputerMethods in Applied Mechanics and Engineering, 55, pp.349-367, 1986.
27. H.L. Wong, "Effects of surface topography and site conditions," Strong GroundMotion Simulation and Earthquake Engineering Applications (R.E.Scholl and J .L.KingEds.), Earthquake Engineering Research Institute, Berkeley, CA, pp.24-1 to 24-5,Nov. 1985.
28. D.E. Beskos, "Boundary element methods in dynamic analysis," Applied MechanicsReviews, 40, No.1, pp.1-23, 1987.
29. F.J. Sanchez-Sesma, "Site effects on strong ground motion," Soil Dynamics andEarthquake Engineering, 6, No.2, pp.124-132, 1987.
30. K.C. Won~, S.K. Datta and A.H. Shah, "Three dimensional motion of a buriedpipeline. I. analysis," J. Engineering Mechanics, ASCE, 112, No. 12, pp.1319-1337,1986.
31. J.E. Luco and J.G. Anderson, "Steady-state response of an elastic half-space toa moving dislocation of finite width," Bull. Seism. Soc. Am., 73, pp.1-22, 1983.
32. A. Mendez and J.E. Luco, "Near-source ground motion from a steady-state dislocation model in a layered half-space," J. Geophysical Research, 93, pp.12041-12054,1988.
33. S.W. Liu, S.K. Datta, M. Bouden and A.H. Shah, "Scattering of obliquely incidentseismic waves by a cylindrical valley in a layered half-space," Earthquake Engineeringand Structural Dynamics, to be published.
34. J.E. Luco, H.L. Wong and F.C.P. de Barros, "Three-dimensional response of acylindrical canyon in a layered half-space," Earthquake Engineering and StructuralDynamics, 19, pp.799-818, 1990.
35. K.F. Graff, Wave Motion in Elastic Solids, Ohio State University Press, 1975.
43
36. J.E. Luco and H.L. Wong, "Seismic Response of Foundations Embedded in aLayered Half-Space," Earthquake Engineering and Structural Dynamics, 15, pp.233247, 1987.
37. J.P. Wolf and P. Obernhuber, "Free-Field Response from Inclined SV- And PWaves And Rayleigh-Waves," Earthquake Engineering and Structural Dynamics, 10,pp.847-869, 1982.
38. A.C. Eringen and E.S. Suhubi, Elastodynamics, Vol. II, Linear Theory, AcademicPress, N.Y., 1975.
39. F. Hartmann, "Computing the C-matrix in non-smooth boundary points," New Developments in Boundary Element Methods, (Ed. by C.A.Brebbia), CML Publications,pp.367-379, 1980.
44
a
A
[Aj
[AJ
b
cp
[C(x)J
d
f
[FJ
NOTATION
numerical coefficient
= ..jk2 - k;= Jk2 - k;measure of incident wave amplitude
square coefficient matrix for the boundary element system (equation32)
3 x 3Mn matrix assembled from the integrals of the products of[FjT and [NJ associated with a single source point (equation 28)numerical coefficients
amplitude of reflected P wave
amplitude of reflected SV wave
numerical constant
= k r (1 - c; / c;)1/2
= kr (1- c;/C~Y/2
P wave velocity including damping
P wave velocity without damping
Rayleigh wave velocity including damping
Rayleigh wave velocity without damping
shear wave velocity including damping
shear wave velocity without damping
3 x 3 matrix related to the geometric shape of the boundary at x(equation 20)
= vi(x - x.)2 + (y - y.)2 + (z - Z.)2, distance between source and receiver points
= J (Y - y.)2 + (z - z.)2, distance between source and receiver pointsprojected on plane x = 0
frequency in Hz
transform€ j full-space Green's traction functions (equation 18b)
45
[F*]
[GJ
k
k'
Ii
L
M
[NJjNi
traction functions associated with full-space Green's displacementfunctions [G*]traction functions associated with the transformed full-spaceGreen's displacement functions [F] for k = a (equation 22b)transformed full-space Green's displacement functions (equation18a)full-space Green's displacement functions
transformed full-space Green's displacement functions for k = a(equation 22a)
=yCI
wavenumber in x-direction
wavenumber in x'-direction
P wavenumber
Rayleigh wavenumber
shear wavenumber
modified Bessel functions of second kind of order zero and orderone
length of the jth element on I'c U I'h
half width of the uniform canyon
discretization range on the half-space surface
number of one-dimensional elements on the boundaries I'c U I'h
number of nodes on the boundaries I'c U I'h
element number for the first and the last one-dimensional elementson boundary l'coutgoing normal to the boundary and its components
shape function matrix for the one-dimensional element on f'c U f'h;
jth shape function
number of elements on canyon boundary l'c
number of elements on one side of the half-space boundary f'h
vector and components of concentrated forces associated with thesecond elastodynamic state
46
{Q}
r.
s
t
{t}
{l}
{t*}
{tf! }
{If f}
{to }
[T]
{u}
{u}
{U/}
{U/}
{u*}
{Uff}
{uff }
{u~ff }
{u. }
{u.}
vector assembled from {Q} associated with all nodes on i'c u f\(equation 33)
3 x 1 vector from integrals of the products of transformed Green'sdisplacement functions and the free-field tractions (equation 29)radius of a circular curve around a source point
the entire boundary of the system which consists of the surface ofthe half-space and the boundary of the canyontime
traction vector
traction vector at plane x = 0
traction vector associated with the second elastodynamic state inreciprocal theoremtraction vector associated with the free-field motion
traction vector associated with the free-field motion at plane x =0
traction vector associated with the scattered field motion
3 x 3 coordinate transformation matrix (equation 6)
displacement vector
displacement vector at plane x = 0
displacement vector in x'y'z'-coordinate system
displacement vector at plane x' = 0 in x'y'zl-coordinate system
displacement vector for the second elastodynamic state in reciprocaltheoremdisplacement vector for the free-field motion
displacement vector for the free-field motion at plane x = 0
displacement vector for the free-field motion in x'y'z'-coordinatesystemdisplacement vector for the free-field motion at plane x' = 0 III
x' y'zl-coordinate systemdisplacement vector for the scattered field
displacement vector for the scattered field at plane x =0
{U}, Uxi , Uyi , Uzi vector containing nodal displacements at plane x = 0 and its components at the jth node
47
x,y,z
x',y',z'
x"' y., z"
'7,.
'7"
spatial coordinates
spatial coordinates in x' y' z'-coordinate system
spatial coordinates of source point
= (x,y,z), an arbitrary point
= (x', y', z'), an arbitrary point In x'y' z'-coordinate system
= (0, y, z), an arbitrary point at plane x = 0
= (O,y.,z.), a source point at plane x=o
number
number
numerical coefficient
canyon sm.'face
cross-section of the canyon at plane x = 0
half-space surface
cross-section of the half-space surface at plane x = 0
extent of the jth element
small contour of radius r. around the source point at plane x = 0
Kronecker delta function
natural coordinate
constant hysteretic damping factor for P wave
constant hysteretic damping factor for Rayleigh wave
constant l:ysteretic damping factor for shear wave
critical incidence angle for SV wave
azimuth incidence angle between the wave propagation plane andx-axis (Figure 2)angle between the normal to the incident wave front and the verticalaxis (Figure 2)shear wavelength
shear modulus for the half-space medium
Poisson's ratio for the half-space medium
48
V'
w
n
wavenumber in zl-direction
material density for the half-space medium
angles of tangent lines passing through a non-smooth boundarypoint on f c U f h
frequency in rad/sec.
non-dimensional frequency defined by n = wL/1rca
49
APPENDIX A: FREE-FIELD GROUND MOTIONS FOR SH, P, SVAND RAYLEIGH WAVES
By definition, the free-field motion is the displacement field that exists inside the
half-space in absence of the canyon. Except for incident Rayleigh surface wave, the
free-field motion consists of the incident wave and waves reflected by the half-space
surface. In the following expressions, the waves are assumed to travel parallel to
x' - Z' plane (see Figure 2). The vertical incidence angle is OIJ and the amplitudes of
the reflected P and S waves are Ap and A. respectively. In X' y'zl-coordinate system,
the free-field displacement fields can be expressed as [35]:
For Incident SH Wave
{uiJ(Xl)} = G) ~',I-'i""'+=""')i + (D .',I-" ••,.'-oo•••• ')i
For Incident P Wave
where
. c..s~n/, = -s~nOIJ
cp
A = sin(201J)sin(2/') - (Cp/C.)2cOS2(2/')p sin(201J)sin(2/') + (Cp/C.)2COS2(2/')
A _ 2(cp/c.)sin(201J)cos(2/')• - sin(201J)sin(2/') + (Cp/C.)2COS2(2/')
For Incident SV Wave
(A'1)
(A·2)
The free-field motions fc,r incident SV wave are given by two different expressions,
depending on whether the vertical incidence angle OIJ is smaller than the critical
incidence angle OIJ defined hy
50
where
• Cp • 0sznj = -szn vC$
A$ = sin(20v )sin(2j) + (Cp/C$)2COS2(20v)
sin(20v )sin(2j) - (Cp/C$)2COS2(Z8v)A
p= sin(20v )sin(2j) + (Cp/C$)2cos2(20v)
where
a = VPsin28 - k2$ v p
-2asin8vsin(20v) + k$ COS2 (20v)iAl = k$ COS2 (28v) - 2asinOvsin(20v)i
k$sin(48v )i
For Incident Rayleigh Wave
(, 2,)e-bL Z _ 1. (2 _ q. )e-bTZ
-. 2 C 'k '{u' (X')} = a • e-' .'"f f , 2,
ik [_h.e-bLZ + -1_(2 - .so. )e-bTZ ]r k~ 2bT C~
where
bL= kr (1 - C;r/2Cp
bT = k r (l- :~r/2
and Cr and k r are Rayleigh wave velocity and the corresponding wavenumber.
51
(A·3)
(A·4)
(A· 5)
APPENDIX B: FULL-SPACE GREEN'S FUNCTIONS
The full-space Green's functions can be found in Reference [38]. Let the source
and receiver points be XOII = (xoI, X0 2, X03) and x = (XI, X2, X3) respectively and n be
the normal to the surface at the receiver x with corresponding components denoted
by n = (nI,n2,n3) where, in order to make the formulation compact, the coordinate
notations have been changed from (x, y, z) to (Xl, X2, X3). The displacement and traction
components at x in the jth direction due to concentrated force at X08 in the kth
direction are given by
where j, k, 1= 1,2,3,
d2 = X2 - X0 2
and kp and k ll are P and S wavenumbers, respectively.
52
APPENDIX C: TRANSFORMED GREEN'S FUNCTIONS
Transformed Green's Displacement Functions
The transformed Green's displacement functions are defined (equation 18a) by
i,j=1,2,3 (C ·1)
where Gij(x,xos) is the ith displacement component at receiver point x = (X17X2,X3)
due to concentrated force at source point xos = (0, X 0 2, X 0 3) in the jth direction in full
space. In order to make the formulations compact, the notation for the coordinates
has been changed from (x,y,z) to (XI,X2,X3). Correspondingly, the coordinates of the
source point become (0, X 0 2, X 0 3).
(C·2)i,j=1,2,3
From Reference [38], the Green's displacement functions, Gij, can be expressed
either in the form given in Appendix B or in the following more compact form
1 [1 [j2 (e- ik•
d- e- ikpd
) 600
0 ]G* - + t3 e-tk• d
ij - 41rp w2 [)Xj[)Xj d de;
where d is the distance between the source and receiver points
In order to obtain explicit expressions for Gij, it is necessary to use following
relation
(C·3)
where b is a complex constant, d is the distance between the source and receIver
projected to the plane Xl = 0:
and K 00 is modified Bessel function of the zeroth order.
Substituting (C· 2) into (C· 1) and utilizing (C· 3) lead to
Gll = 2~Jl [Ko(asd) (1 - ~;) + ~;Ko(apd)] (C ·4)
53
where
ap = Jk2 - k~
as = y'k2 - k;
j = 2,3
a,f3 = 2,3
(C·5)
(C·6)
and J1 is the shear modulus for the full-space medium. It should be pointed out that
k is a real variable while ap and as are complex numbers in general. When damping
is included, ap and as never vanish. Thus, Gij is always bounded for J =I o.
Transformed Green's Traction Functions
nThe transformed Green's traction functions Fij(xo, xos, k) are defined (equation
18b) by
i,j = 1,2,3 (C·7)
nwhere Fij(x,xos) is the ith traction component at the receiver point x = (X},X2,X3) due
nto a concentrated force at source point X08 = (0, X 0 2, X 0 3) in jth direction. Fij can be
nobtained by substituting the expressions of Ftj' which are given in Appendix B, into
(C . 7) and finding the integrations term by term. However, since finding derivativesn
is much easier than deriving integrations, it is more convenient to find Fij through
following ways:
(1) Assume a displacement field given by
G ( - - k) ikxlUij = ij x o, Xos, e (C·8)
(2) Find the corresponding stresses and tractions for this field and express the
latter as
(C·9)
n(3) Remove eikx1 from (C· 9) and one obtains Fjj
54
(C ·13)
nIt can be shown that Fij thus obtained are the same as what would be obtained
nby substituting Frj into equation (C· 7) and deriving the integrations term by term.
nThe final expressions for Fij are given by (j = 1,2,3)
F1j = Jl [n2 (88~:j + ikG2j ) + n3 (88~~j + ikG3j) ] (C· 10)
F2j = Jl{ 1~~v [ikVGlj + (1 - v) 8:X:j+ v 88~:j ] + n3 (88~:j + 88~:j) } (C ·11)
F3j = Jl{ 1~~v [ikVGlj + (1 - v) 8:X:j+ v 8:X:
j] + n2 (88~:j + 88~:j) } (C ·12)
where n2 and n3 are the components of outgoing normal to the surface, n = (0, n2, n3),
and v is the Poisson's ratio for the half-space medium.
Expression for Matrix [C]
Matrix [C] depends on the position of the source point; and, when the source
point is on the boundary, it is a function of the geometric shape of the system at
the source point. From equation (20), matrix [C] is defined by
[C(xos)] = lim f [F(xo,xos,k)fdrr.-o}r.
nSubstituting [F] into (C·13) and following similar procedures presented in Reference
[39], it can be found that
(C ·14)
where
1Cll = 21r (</>1 - </>2) (C ·15)
C22 = 41r(11_ v) [2(1- V)(</>1 - </>2) + %(Sin(2</>d - Sin(2</>2»] (C ·16)
C23 = C32 = (1 ) (sin2</>1 - sin2</>2) (C ·17)41r 1 - V
C33 = 41r(11_ v) [2(1- V)(</>1 - </>2) - %(sin(2</>d - Sin(2¢>2»] (C .18)
In the above equations, </>1 and </>2 are the angles of tangent lines passing through the
source point xos at the boundary. The detailed definition of </>1 and </>2 can be found
in Reference [39].
55
PART II
IMPEDANCE FUNCTIONS FOR THREE-DIMENSIONAL
FOUNDATIONS SUPPORTED ON AN INFINITELY
LONG CANYON OF UNIFORM CROSS-SECTION IN A
HOMOGENEOUS HALF-SPACE
INTRODUCTION
Required in the substructure method for earthquake analysis of concrete dams is
the impedance matrix (or the frequency-dependent stiffness matrix) for the foundation
rock region, defined at the nodal points on the dam-foundation rock interface [1,2].
Computation of this foundation impedance matrix for analysis of arch dams requires
solution of a series of mixed boundary value problems governing the steady-state
response of the canyon cut in a three-dimensional, unbounded foundation rock region.
Because such analyses are extremely complicated, usually only the foundation flexibility
is considered in analysis of arch dams, i.e. the material and radiation damping as well
as inertial effects of the foundation rock are ignored [2-4]. The objective of this work
is to overcome these limitations with the long term goal of analyzing the earthquake
response of arch dams including dam-foundation rock interaction.
Computation of foundation impedance matrices has been the subject of many
investigations over the past two decades leading to a variety of analysis techniques.
Impedance matrices for two-dimensional embedded foundations have been determined
by the indirect boundary element method (BEM) with half-space Green's functions
[5], direct boundary element method with full-space Green's functions [6], and by a
hybrid method [7]. Procedures were developed to determine the impedance matrices
for three-dimensional surface-supported or embedded foundations by an integral equa
tion technique with half-space Green's functions [8,9], nonsingular integral equation
approach [10-12], and a hybrid method combining indirect BEM with FEM using
half-space Green's functions [13].
Boundary methods utilizing half-space Green's functions have the advantage that
they are applicable to layered media and do not require discretization of the half-space
surface. However, because of the high cost in computing half-space Green's functions,
application of these methods has been restricted to either arbitrary-shape, surface
supported foundations or axisymmetric embedded foundations. In contrast, full-space
59
Green's functions are simple to compute and have been used widely in the direct
BEM to determine foundation impedance matrices [14-19]. However, use of full-space
Green's functions seems computationally impractical for analysis of foundations In
layered media because of the requirement to discretize interfaces between layers.
In addition to various boundary methods using different Green's functions, other
techniques have been developed for evaluation of foundation impedance matrices.
These include the finite element method in the frequency domain with transmitting
boundaries [20-23], finite element method in the time domain [24], spherical wave
function expansion approach [25], infinite element approach [26,27] and various hybrid
methods [e.g. 28,29]. More extensive reviews of the various techniques developed to
determine foundation impedance matrices are available in [30,31].
The various techniques mentioned above have been applied to the analysis of
a wide variety of foundations either supported on the surface of a half-space or
embedded In a finite-sized excavation in the half-space. However, these techniques
appear to be either inapplicable or computationally impractical for the system of
Figure 1. It is an infinitely-long canyon of arbitrary but uniform cross-section cut
in a homogeneous viscoelastic half-space. In this work this system is analyzed to
determine the impedance matrix at the arch dam-foundation rock interface by the
direct boundary element method in conjunction with full-space Green's functions. The
uniform cross-section of the infinitely-long canyon, which seems to be a reasonable
idealization for practical analysis of arch dams, permits analytical integration along the
canyon axis leading to a series of two-dimensional boundary problems involving Fourier
transforms of the full-space Green's functions. The assumption of a homogeneous half
space seems to be appropriate for arch dam sites where similar rock usually extends
to considerable depth.
PROBLEM STATEMENT
The system considered is an arch dam, idealized as a finite element system,
60
supported in an infinitely-long canyon of arbitrary but uniform cross-section cut in
a homogeneous half-space (Figure 1). The substructure method has proven to be
an effective approach to the earthquake analysis of dams including dam-foundation
rock interaction effects. These interaction effects introduce an impedance matrix, or
a dynamic (frequency-dependent) stiffness matrix, for the foundation rock region in
the equations of motion governing the steady-state response of the dam to harmonic
ground motion.
Our objective IS to determine the impedance matrix, [SAw)], where W IS the
frequency of the harmonic excitation, defined at the interface ri between the dam
and foundation rock. This impedance matrix, [Sf(w)], relates the interaction forces
{R(t)} at the interface ri to the corresponding displacements {ret)}, relative to the
earthquake induced ground displacements in the absence of the dam:
[SJ(w)]{r(w)} = {R(w)} (1)
where the overbar denotes Fourier transform of the time-functions. The square matrix
[Sf(w)] is of order equal to the number of degrees-of-freedom (DOF) in the finite
element idealization at the interface. The nth column of this matrix multiplied by
eiwt is the set of complex-valued forces required at the interface DOF to maintain
a unit harmonic displacement, eiwt , in the nth DOF with zero displacements in all
other DOF.
Evaluation of these forces requires solution of a senes of mixed boundary value
problems with displacements prescribed at the interface ri and tractions outside ri
- on the canyon as well as the half-space surface - are prescribed as zero. Instead
of directly solving this mixed boundary value problem, it is more convenient to solve
a stress boundary value problem in which non-zero tractions are specified at the
interface r i and the resulting displacements at r i are determined. Assembled from
these displacements, the dynamic flexibility influence matrix is inverted to determine
the matrix [Sfew )].
61
Fig.! Infinitely-long canyon of arbitrary but uniform cross-section; rc is thecross-section of the canyon at x = 0; r i is the interface between thedam and foundation rock.
62
The frequency-dependent impedance matrix [S/(w)] is function of the material
properties of the foundation rock and the geometry of both the canyon and the
dam-foundation rock interface. For the system considered, the material properties
of the half-space medium are characterized by the complex P and S wave velocities
Cp =cp(l + i77p)1/2 and Cs = cs(1 + i77s)1/2. The terms 77p and 77s represent the constant
hysteretic damping factors for P and S waves respectively; and cp and Cs denote the
P and S wave velocities if the medium were undamped. Similarly, the Rayleigh wave
velocity is given by Cr = cr(l + i77r)1/2 where Cs is the Rayleigh wave velocity without
damping and 77r is the corresponding hysteretic damping factor. Both Cr and 77r are
related to cp, cs , 77p and 778'
FOURIER REPRESENTATION OF TRACTIONS AND DISPLACEMENTS
Because the system geometry is uniform along the length of the canyon, it is
useful to express tractions and displacements through their Fourier integral represen
tations. The tractions are prescribed over the dam-foundation rock interface, which is
discretized into two-dimensional curved boundary elements compatible with the finite
element discretization at the base of the dam. The prescribed traction on the jth
boundary element is {t(x)}jeiwt , where x = (x,y,z) is an arbitrary point within the
element, and the vector {t(x)}j contains the x, y and z components of tractions. The
traction vector is expressed through its Fourier integral representation:
(2)
where Xo = (0, y, z) and {i(io , k)}j is the Fourier transform of {t(i)}j with respect to x
(3)
The Fourier transform parameter k can be interpreted physically as a wavenumber
since {i( i o , k)} je- ikxeiwt represents tractions associated with a wave travelling along
the x-axis. Thus equation (2) suggests that the prescribed tractions can be considered
as a superposition over all wavenumbers of tractions associated with waves travelling
63
in opposite directions along the x-axis. In particular, a negative value of k indicates
waves travelling in the negative x-direction and a positive k represents waves moving
in the positive x-direction. The superposition of these waves, which travel at speed
c =w/lkl, results in the prescribed tractions in the form of a standing wave.
The prescribed tractions on the dam-foundation rock interface are next expressed
in terms of traction distributions over individual boundary elements. Along the line
parallel to the x-axis passing through xo , the traction distribution is given by the
superposition of tractions associated with each of the elements intersected by the line:
where
{i(xo , k)} = I)i(xo , k)}jj
(4)
(5)
The tractions are obviously zero at all locations on the canyon surface and half-space
surface that are outside the dam-foundation rock interface.
The steady-state displacements {u(x) }eiwt resulting from the prescribed tractions
with harmonic time-variation at frequency w can also be expressed as a superposition
of displacements associated with travelling waves. Thus, the vector {u( x)} containing
the three components of displacements at the canyon boundary r c, or the half-space
surface rh, resulting from the prescribed tractions (equation 4) is given by
(6)
where {u(xo , k)} is the Fourier transform of the displacement vector gIven by
{U(xo , k)} = 21 joo {u(x)}eikxdx7r -00
(7)
For any particular wavenumber k, {u(xo , k)}e- ikx represents a wave with displacement
pattern {u(xo , k)} travelling in the x-direction with speed c = w/lkl. As mentioned
before, a positive value of k indicates a wave travelling in the positive x-direction,
64
and vice-versa. In addition, the dependence of the displacements and tractions on
the frequency w is implied throughout.
We have now expressed the prescribed' tractions and unknown displacements
In terms of their Fourier integral representation which implies superposition over
all wavenumbers (equations 4 and 6). Considering one wavenumber at a time, the
prescribed boundary tractions are {t(xo,kne-ikxeiwt and the unknown displacements
are {u(xo,kne-ikxeiwt, where w is the excitation frequency and k is the wavenumber.
The problem to be solved for a particular wavenumber involves determining the
displacements {u( i o , kn at the curve f' c, where x = 0, due to the travelling wave
associated with tractions {t(io,kn obtained from equation (5) at x = o. The resulting
displacements for all wavenumbers are superposed to obtain the displacements at
any location i on the boundary (equation 6); in particular, the displacements at the
dam-foundation rock interface, f i , can be determined.
BOUNDARY INTEGRAL FORMULATION
The boundary integral equation for the system considered, consisting of an in
finitely long canyon of uniform cross-section, can be obtained from the reciprocal
theorem and can be expressed as (Part I)
(8)
where {t(in is the prescribed traction at i on the boundary with its normal denoted byn
n, {u(xn is the resulting unknown displacement vector, and [F*(i, ios)] and [G*(i, ios)]
are 3 x 3 matrices of Green's functions for a full-space. The first, second and third
columns of these matrices correspond to the traction and displacement vectors at i
due to a unit point load applied in the x, y and z directions, respectively, at a source
point ios = (0, ys, zs) on f' c or f'h, the intersection line of the system boundary and
the plane x = o. The integration is over the entire system boundary, consisting of the
surface of the half-space and the boundary of the canyon, modified slightly to avoid
the singularity in the Green's functions at the point of load application (Figure 2).
65
Fig.2 Definition of boundaries and source point; rc is the cross-section ofthe canyon at x = 0, rh is the x = 0 line on the surface of the halfspace, and r. is the small contour of radius T. around the sourcepoint.
66
Expressions for [F*(x, xos)] and [G*(x, xos)] are given III Appendix B of Part I. For
each source point, xos, equation (8) provides three scalar equations in the unknown
displacements {u( x)}.
Substituting equations (4) and (6) into equation (8) and interchanging the order
of integration lead to
=0
After analytically evaluating the integral along the x-axis, equation (9) becomes
(9)
where
(lla)
(llb)[F(xo,xos,k)] = i:[F*(fi,xos)]e-ikXdx
Expressions for the transformed Green's functions, [G] and [F], are given in Appendix
C of Part I.
As mentioned earlier, the displacements resulting from prescribed tractions
{f(xo,k)}e-ikxeiwt have the form {u(xo,k)}e-ikxeiwt, i.e., {f(xo,k)} and {u(xo,k)} corre
spond to the same wave travelling at speed c = w/k, where k is the wavenumber.
Thus, equation (10) implies that the following equation must be satisfied for each k
The final step is to make the radius rs of the small arc f's around the source
point xos approach zero. Following the steps in [32], the final form of the boundary
integral equation (12) can be expressed as
67
(13)
where
[C(xos)] = lim f [F(xo, Xos, k)]Tdr (14)r.-O Jr.
The general expression for the 3 x 3 matrix [C(xos)] is given in Appendix C of Part
I. It is independent of the wavenumber k and depends only on the geometry of the
boundary at the source point X08" If the boundary is smooth, i.e., it has a unique
tangent, Cij(ios) = 6ij /2.
Equation (13) is the exact formulation for the problem, requiring solution of the
displacements {u(xo, k)} along f\ and f\, i.e. x = 0, due to tractions {f(io, k)}e- ikx
associated with a single wave travelling at speed c = w/k. In this equation, {f(xo,k)}
is known from equation (5), the transformed Green's function matrices [G] and [F]
are known from equation (11) and Appendix C of Part I, whereas the unknowns
are the displacements {u(xo,k)} at both the source point xos and the rest of the
boundary f\ u rh. Thus, by taking advantage of the uniform geometry of the system
along the x-direction, the displacements associated with a single travelling wave can
be determined by solving a "two-dimensional" boundary integral problem (equation
13) in which the remaining integration is only along the curve r c U rh •
Solution of an infinite number of such two-dimensional problems, each corre
sponding to a particular wavenumber, would lead to the displacements {u(xo , k)} for
all wavenumbers, which are combined in accordance with equation (6) to determine
the displacements {u( i)} at any location on the boundary, and in particular at the
dam-foundation rock interface. In practice, of course, the infinite integration range
in equation (6) would be replaced by a finite range and the truncated integral would
be numerically evaluated from results at an appropriate set of wavenumbers.
BOUNDARY ELEMENT FORMULATION
Discretization
In its present form, equation (13) cannot be solved analytically for the displace-
68
ments {u( X0, k)}, even for the simplest canyon geometry. Therefore, the integration
domain feU f h is discretized into M one-dimensional line elements (Figure 3). Along
the half-space surface f h , the discretization obviously extends to only a finite dis
tance L y , which should be large enough to provide accurate solutions. The number
of nodes in each element depends on the element type; in this study, two node el
ements are used with linear interpolation functions given by N1(O = 0.5(1 - () and
N2(O = 0.5(1 + 0 where ( varies from -1 to 1. Thus the discretized system with M
elements has M n = M +1 nodes.
The variation of the unknown displacements {u(xo , k)} over the jth element on
feu f h can be expressed in terms of nodal displacements
where [N(O] is a 3 x 6 matrix consisting of interpolation functions
(15)
(
NI[N(O] = ~
N2 0o N 2
o 0(16)
and {Uh IS the vector of displacements at the two nodes - j and j + 1 - of the
element
As mentioned before, the dam-foundation rock interface f i IS discretized into
two-dimensional curved elements compatible with the finite element discretization at
the base of the dam. The total number of nodes on r i is denoted by Mi. On each
element, the prescribed tractions can be expressed in terms of nodal tractions; for
the jth element
(17)
where the matrix [h(x))j consists of two-dimensional interpolation functions and {T}j
is the vector of tractions at all the nodal points of the element. If the jth element
has M nodes, [h(x)]j is of size 3 x 3M and {TL is of dimension 3M. As a result, the
69
x
Fig.3 Boundary element discretization.
70
y
y
Fourier transforms of the prescribed tractions (equation 3) in the jth element on fj
can be represented by
(18)
where [h(xo,k)]j is the Fourier transform of the two-dimensional interpolation functions
with respect to x
(19)
Expressions of [h(xo,k)L for 4, 6 and 8 node elements are given in Appendix A.
Combining the contributions of all boundary elements on f j intersected by the
line x= Xo (equation 5), the Fourier transform of the traction distribution along this
line is given by
{t(xo,k)} = L [h(xo,k)]j{T}jj
a"'''emble
which, after assembly, can be expressed in matrix form as
(20)
(21)
In equation (21), {Tho contains the tractions at all the nodal points on the elements
intersected by the line x= xo •
Thus, after discretizing l'c ul'h into one-dimensional line elements and f j into two
dimensional elements, the Fourier transforms of the unknown displacements, {u(xo , k)},
have been expressed in terms of nodal displacements (equation 15) and, {t(xo,k)}, the
Fourier transforms of the prescribed tractions, have been represented through nodal
tractions {T} (equation 21).
Boundary Element Equations
The next step is to replace the boundary integrations III equation (13) by the
summation of integrals over all the M elements
M
[C(xos)]{u(xos, k)} +L 1[F(Xo, Xos, k)]] {u(xo,k)}jdfj=l t j
71
M2 .
= L 1[G(Xo,xos,k)];{t(xo,k)}jdrj=M1 fj
(22)
where rj is the extent of the jth element. The summation on the right side only
covers the elements on the canyon boundary, rc, since the tractions are zero on the
half-space surface, rh. By expressing the displacements and the tractions in terms
of nodal displacements and tractions through equations (15) and (21) and changing
the integration variable to (, the contributions from the jth element in equation (22)
become
~ [G(xo, xos, k)]; {[(xo, k)}jdr = ~Ij11
[G((, Xos, k)]J[Ji((, k)]j{T},d( (23b)if j 2 -1
where Ij is the length of the jth element. Substituting these equalities, equation (22)
can be expressed in matrix form
(24)
Equation (24) represents three equations associated with each source point xos =
(0, Ys, zs) and each wavenumber k, where the 3 x 3 matrix [C(xos)] is given by equation
(14), the 3 x 1 vector {u(xos,k)} represents the unknown displacements at the source
point, {iT} is a 3Mn x 1 column vector of unknown nodal displacements (3 displacement
components at each of the Mn nodes on the boundary rcurh ), {T} is a 3Mi X 1 column
vector of nodal tractions (3 components at each of the M j nodes on r j ), [A(xos, k)]
is a 3 x 3Mn matrix assembled from the integrals of the products of the transformed
traction Green's functions and element interpolation functions
- 1[A(xos, k)] == 2 (25)
and [B(xos,k)] is a 3 x 3Mi matrix assembled from the integrals of the products of
the transformed displacement Green's functions and Fourier transform of the two-
72
dimensional interpolation functions (equation 21) ,
- _ 1 M 2 II _ T -[B(xos,k)] = - L lj [G((,xos,k)]j [H((,k)]jd(
2 ;=Ml -IG,uemole
(26)
Both [A(xos, k)] and [.8(xos, k)] are obtained by appropriately assembling the contri
bution of the various elements, a process similar to the stiffness matrix assembly in
the direct stiffness method. The construction of both of these matrices for a small
system is illustrated in Figure 4.
Equation (24) is a discrete form of equation (13) obtained by introducing two
approximations: (1) interpolation of the unknown displacements over each element on
l"curh in terms of nodal displacements (in this particular formulation, the interpolation
functions were chosen to be linear); and (2) truncation of the integration on rh to a
finite distance.
Formation of Systenl Matrices
Associated with the source point xos = (0, Ys, zs), equation (24) represents 3 linear
algebraic equations in 3Mn unknown displacements. Clearly, a total of 3Mn equations
are needed to determine the unknown displacements. These equations can be obtained
by successively choosing the source point at each of the nodes of the discretized
boundary r c Urh. When the source point coincides with the ith node, equation (24)
becomes
(27)
By varying i = 1,2, ... , Mn and assembling the resulting equations, a system of 3Mn
linear equations can be obtained
[A(k)]{U} = [B(k)]{T}
where [A(k)] is a square matrix of order 3Mn , given by
(
[C(XI)] ) ([A(XI'k)])
[A(k)] = + .
[C(XMJ] [A(x~n,k)]
73
(28)
(29)
x
2
7 y
contribution ofelement 3 = [A(xo""k)h
s
contribution ofelement 3 = [B(xo""k)h
}{T}~==c;..;:.==c;..;:.="'=";;.I=.==...=..;;J
1•
Ufc[C(xos)]{u(xos, k)} +{
~~~~~
+
FigA Assembly of Matrices.
74
and [B(k)] is a 3Mn x 3Mi matrix given by
(
[B(.:l' k)J )
[B(k)] = _
[B(XMn ' k)J
(30)
The coefficient matrix [A(k)J is fully populated and unsymmetric, which is of typical
of direct boundary element method equations.
Evaluation of Displacements
Solution of the equation (28) leads to the displacements {tI} at the M n nodes on
the boundary f' c u f' h in terms of the nodal tractions {T} for a particular wavenumber
k
{tI} = [A(k)t1[B(k)]{T} (31)
Once the nodal displacements are determined, the displacement variation {u(x0' k)}
over each finite element on f' c can be obtained from equation (15).
IMPEDANCE MATRIX
After the displacements {u(xo , k)} at the y-z plane (x = 0) of the canyon boundary,
f' c, and the half-space surface, f'h, are obtained for various values of the wavenumber,
k, the displacements at any other location on these boundaries can be obtained from
equation (6). In particular, the displacements at the dam-foundation rock interface,
ri, can be determined. The indefinite integral of equation (6) is truncated at a finite
range (-K, K) and it is evaluated from the displacements obtained for a discrete set
of wavenumbers k q using an appropriate integration scheme:
{u(X)} = L Wq{u(xo , kq)}e-ikqxq
(32)
where Wq are weighting coefficients associated with the numerical integration scheme,
and kq is the qth sampling point in the wavenumber domain k. Guidelines for selecting
sampling points in the wavenumber domain will be presented later.
75
Thus, the vector of displacements at the jth node on fi, defined by the coordinates
Xj = (x j, Yj, Zj), due to the prescribed tractions is given by
{-} "W{-(- k)}-ikqxjr j = L.J q U Xoj, q eq
(33)
where i oj = (0, Yj, Zj), and {u(xoj,kq)} is obtained from the solution of equation (31)
for the nodal displacements and the displacement interpolation relation (15). If Xoj
falls in the mth element (Figure 3) in the plane x = 0 with (j as the corresponding
natural coordinate, {u(xoj,kq )} is given by (see equation 15)
(34)
where N 1((j) and N 2((j) are the linear interpolation functions given in the previous
section, and {U}m and {U}m+l are the 3 x 1 vectors of displacements at the mth and
(m+l)th nodes on fcufh associated with the qth sampling point k q • Since the nodal
displacements are linear combinations of the nodal tractions (equation 31), {u(xoj,kq )}
in equation (34) can also be expressed as linear combinations of {T}, leading to
(35)
where [E(xoj,kq)] is a 3x3Mi matrix relating the Fourier transform of the displacements
along the line x= Xoj to the nodal tractions. Thus from equations (33) and (35), the
vector of displacements at the jth node on f i becomes
where
[jj] = L Wq[E(xoj, kq)]e-ikqxjq
Combining equations (36) for each of the Mi nodes on fi leads to
{f} = [f]{T}
76
(36)
(37)
(38)
where {r} is a 3Mi x 1 vector of nodal displacements ( 3 components at each of the
Mi nodes) and [f] is a square matrix of order 3Mi given by
(
[id )[I] = _:
[lMi]
(39)
The matrix [f] in the above equation is the frequency-dependent flexibility influence
matrix which relates the displacements at the DOF on the dam-foundation rock
interface fi to the corresponding nodal tractions {T}.
As mentioned earlier, the impedance matrix [Sj(w)] for the foundation rock relates
nodal forces to nodal displacements at the discretized dam-foundation rock interface.
This matrix is obtained from equation (38) by transforming distributed tractions to
nodal forces:
{R} = [~]{T} (40)
where {R} is a 3Mi x 1 vector of nodal forces ( 3 components at each of the Mi nodes
on f i ), and [~] is a square matrix of order 3Mi assembled from the products of the
two-dimensional interpolation functions
[~] =Ii [h(x)f[h(x)]df
Combining equations (38) and (40) gives
{R(w)} = I~][I(w)]-l{r(w)}
(41)
(42)
wherein it is recognized that the nodal forces {R(w)}, the nodal displacements {few)}
and the flexibility influence matrix [few)] are frequency-dependent. Thus, comparing
equations (1) and (42) leads to the 3Mi x 3M; impedance matrix
[Sew)] = [~][f(w)]-l (43)
The impedance matrix determined from equation (43) will not be exactly sym
metric due to the approximations inherent in the boundary element procedures; it is
77
nearly symmetric, however. By minimizing the 'differences in the unsymmetric off
diagonal terms in a least square sense [16], a symmetric impedance matrix is given
by
(44)
For a rigid foundation on the surface of the canyon, the nodal displacements at
the foundation-rock interface can be expressed as
{r(w)} = [a]{ro(w)} (45)
where {ro(w)} is the 6 x 1 vector containing the six rigid-body DOFs of the foundation
and [a] is a 3Mi x 6 matrix relating the nodal displacements on r i with {ro(w)}. Thus
the 6 x 6 impedance matrix for the rigid foundation is given by
[Sr(w)] = [a]T[Sj(w)][a]
SUMMARY OF PROCEDURE
(46)
The boundary element procedure developed in the preceding sections to determine
the impedance matrix, or dynamic stiffness matrix, corresponding to an excitation
frequency w, for the foundation rock region, defined at the DOF of the dam-foundation
rock interface, may be summarized as follows:
1. Discretize ri, the dam-foundation rock interface, into two-dimensional surface ele
ments, select interpolation functions [hex)], and determine their Fourier transforms
(equation 19). The expressions of these Fourier transforms for 4, 6 and 8 node
surface elements are available in Appendix A.
2. Discretize fe, the canyon boundary at x = 0, and f h, the half-space surface at x = 0,
into M line elements covering a range Lyon fh (Figure 3). If linear interpolation
functions are selected for each element, the discretized system contains M n = M +1
nodes.
3. Choose an appropriate integration scheme and its associated weighting coefficients
Wq and a discrete set of wavenumber sampling points kq (equation 32). Guidelines
78
for selection of kq and the truncated integration range (-K,K) are presented in
a later section. For each k = kq , repeat steps 4 to 7 to express the nodal
displacements {U} on f'cuf\ in terms of nodal tractions {T} on ri (equation 31).
4. Establish equation (27) which includes three linear algebraic equations associated
with Xi, the ith node on f' c U f'h' in 3Mn unknowns (3 displacement components
at each of the M n nodes) by the following steps:
(a) Compute the elements of the 3 x 3 matrix [C(Xi)] which depend on the geometry
of the boundary at the ith node; see Appendix C of Part 1. These elements
are given by Cjk(Xi) = bjk/2 if the boundary is smooth, i.e., it has a unique
tangent at Xi.
(b) Compute the 3 x 3Mn matrix [A(Xi, kq )] from equation (25) by assembling the
contributions of all the M elements, as described in Figure 4. Element contri
butions are determined from the integrals of the transformed traction Green'sn
functions, [F], (equation 11b) and element interpolation functions. Analyticaln
expressions for [F] are available in Appendix C of Part 1.
(c) Compute the 3 x 3Mi matrix [B(Xi, kq )} from equation (26) by assembling the
contributions from all the elements on f'c, as described in Figure 4. The
line element contributions are determined from equation (26) wherein the
transformed displacement Green's functions, [GJ, are computed from equation
(l1a) and the assembled Fourier transforms of the two-dimensional interpolation
functions, [ill, from equation (21). Analytical expression for [G] is available
in Appendix C of Part I and the Fourier transforms of the t\Yo...dimensional
interpolation functions are given in Appendix A.
5. Repeat the computations summarized in step 4 for each of the nodes, i = 1,2, ... , Mn
to determine [C(Xi)], [A(Xi' kq )] and [B(Xi' kq )] for all the nodes.
6. Evaluate the square matrix [A(kq )] of order 3Mn and the rectangular matrix
[B(kq )] of size 3Mn x 3Mi for the boundary element system from the corresponding
79
individual nodal matrices (step 5) using equations (29) and (30).
7. Solve the system of 3Mn linear algebraic equations (28) to express the nodal
displacements {U} on f'c u f'h in terms of the nodal tractions {T} (equation 31).
8. For each sampling point k = kq , compute the 3 x 3Mi matrix [E(xoj, kq )] (equation
35) by interpolating the nodal displacements {U} determined in step 7.
9. Compute the 3 x 3Mi matrix flj(w)] from equation (37) wherein [E(xoj, kq )] for any
k q is known from step 8 and the wavenumber sampling points k q were selected
in step 3.
10. Repeat the computations summarized in step 9 for each of the nodes, j = 1,2, ..., Mi'
on fi and combine the resulting [jj(w)] according to equation (39) to obtain the
3Mi x 3Mi flexibility influence matrix [j(w)].
11. Compute the 3Mi x 3Mi transformation matrix [~] from equation (41) wherein
[h(x)] was determined in step 1.
12. Compute the square matrix [Sew)] of order 3Mi from equation (43) wherein [jew)]
and [~] are known from steps 10 and 11. The symmetric impedance matrix [Bj(w)]
is determined from equation (44).
13. Repeat the computations of steps 3 through 12 for each harmonic excitation
frequency w of interest to obtain the corresponding impedance matrix. The ex
citation frequencies should be selected to cover the frequency range over which
the ground motion and dam response are significant.
NUMERICAL ASPECTS
Selection of Sampling Points in Wavenumber Domain
As mentioned in steps 3, 9 and 10 of the procedure summary, the computation to
determine the nodal displacements {U} is repeated for a discrete set of wavenumbers
kq distributed over a truncated range (-J(, J(). However, analyses are necessary only
for positive wavenumbers because they indirectly also provide the displacements as-
80
sociated with corresponding negative wavenumbers·(Appendix B). Having determined
the displacements {il} associated with various kq and {u(xo , kq )} from equation (34),
the displacements at the dam-foundation rock interface fi are determined by numeri
cally evaluating the integral of equation (6). Since solution of the "two-dimensional"
problem to determine {u(xo,k q)} requires much more computational effort compared
to evaluation of the integral of equation (6), the integration scheme chosen should
require as few sampling points in k q as sufficient for required accuracy.
In order to properly select the discrete set of wavenumbers over the range (0, K),
the variation of {u(xo , k)} with k is examined next. This variation is plotted for the
system shown in Figure 5. It is an infinitely-long canyon with a semi-circular cross
section of radius L and a dam-foundation rock interface of width b = 0.2L subjected
to uniform horizontal traction on the interface. The displacement-wavenumber plot
for point A is presented in Figure 5. For each wavenumber k, the displacement at
A is determined by implementing steps 4 to 7 of the procedure summary wherein
{T}T = (0,1,0, ... ,0,1,0) in equation (31). It is apparent that the displacements peak
sharply at the Rayleigh wavenumber: kr =l.07 for ao = 1 and kr = 4.29 for ao = 4
where a o = wLlcs is the normalized frequency. Thus finely-spaced wavenumber values
would be necessary in the vicinity of kr, whereas coarse spacing would suffice away
from k r •
Thus the wavenumber domain (0, K) is divided into several subdomains, each
with different step size, with the subdomain as well as step sizes depending on the
excitation frequency, and the size and discretization of the system. Typically the
four subdomains chosen were: (0, kr - L1k), (k r - L1k, kr + L1k), (k r + L1k, kr + L1k1 ) and
(k r +L1k1 ,IO where °< L1k < kr and L1k1 ~ 3L1k. The integration is truncated at kq = K
when the maximum amplitude of the displacements {D} is smaller than a specified
tolerance. The four-point Gauss integration scheme utilized is convenient to implement
and flexible to use. However, it has the disadvantage that, when accuracy needs to
be improved, the previous results cannot be reused and the entire computation has
81
yL L
_--- uniform horizontaltractions on r i
A
---1\
y
z
0.15
0.10
I- ,0 0 =.::L ~
0 II
tx II......... ,I1::3 ,I
,I
0.05 III,,
,J
0.00
I,IIIIIII
II 00
= 4II
",I,II I, I, I
I II I, I, I
I II I
, I, I
, I
,'~ ".. .. \ ,, ...
------10.1----------·10 I- - - - - . IO~I
o 2kr = 1.07
4 kr =4.29 6Wovenumber k
8 10
Fig.5 Variation of displacements {u(xo , k)} at A with wavenumber kj a o =wL/cs = 1 and 4, v = 1/3, .1]p = 1]s = 0.02.
82
to be repeated for a new set of wavenumbers.
Alternately, more complicated adaptive integration schemes may be employed to
permit automatic control of errors and reuse of results from previous calculation. Since
the actual integrand for the present problem is not one function but a matrix function
with many entries (equation 37), direct implementation of an adaptive procedure may
be inconvenient because large computer storage or input-output operation is needed
to save the previously computed results. Therefore, the discrete set of wavenumber
sampling points, kq , are determined from a simple test function, which has the main
features of the integrand.
From equation (31) and Appendix B, the nodal displacements on the boundary
I'cuI'h can be expressed as
(47)
where [Bc(k)] consists of the transformed displacement Green's functions, [G], and
[BH(k)] consists of the Fourier transforms of the two-dimensional interpolation func
tions (equations 19 and 21), [H(xo,k)]. The simple test function may be obtained
by analyzing the main feature of each of the three matrices, i.e. [A(k)]-l, [Bc(k)]
and [BH(k)], in equation (47). First, the entries in [A(k)]-l have peaks of different
amplitudes near the Rayleigh wavenumber. Thus, the wavenumber dependence of
[A(k)]-l may be characterized by l/g(k) where g(k) is the Rayleigh function given by
[33]
(48)
Secondly, as k becomes greater than kr and further increases, the nodal displacements
{U} decrease which, as will be seen in the next subsection, is primarily controlled
by the asymptotic behavior of the modified Bessel functions Ko(asd) and K1(asd) in
[Bc(k)]. The slowest decay among all elements in [Bc(k)] with respect to k may be
controlled by KI(asdmin) where dmin is the shortest of all distances between a source
point, xos, and all the integration points on I'c U I'h used to evaluate equation (26).
83
Finally, the high frequency content in the variation of the nodal displacements {tT}
with respect to k is mostly due to the exponential terms in the matrix [BH(k)]. This
variation should be no faster than eiklxlmu where Ixl max is the maximum absolute value
of the x-coordinate for the nodes on the dam-foundation rock interface (Appendix A).
Combining the main features of the three matrices in equation (47), a test function
may be chosen as
4>(k) = J(1(;(:)in) eiklxlmu (49)
It should be pointed out that the 4>(k) of equation (49) does not represent any term
in the matrix [A(k)t 1 [Ba(k)][BH(k)] but is intended to represent the main variation
features of the matrix with respect to k. In particular, for a system with larger dmin
and Ixl max , equation (49) implies that finer sampling points over the wavenumber
range (0, J() should be selected, but the integration also can be truncated at a shorter
J(.
Based on this test function, any adaptive integration scheme, such as that based on
Chebychev approximation [34] or quartic polynomial interpolation [8], may be used
to determine the wavenumber sampling points, kq • In this investigation, a simple
adaptive integration scheme based on Simpson's rule was applied to the test function
to determine k q • In this scheme, the integration of the test function over a typical
interval, (k1 , k2 ), is first computed by
(50)
Then the integration is computed again by inserting two more sampling points midway
between the first three sampling points through
12 = (k2;;k1
) [4>(k1 ) +44>(3k1: k2
) + 24>( k1; k2
) + 44>( k1: 3k2
) +4>(k2 )] (51)
If the relative error between II and 12 is below a specified tolerance, the five sampling
points for the interval (kI, k2 ) are saved for later use. Otherwise, the process is
repeated for the two intervals (kI, (k1 + k2 )/2) and ((k1 + k2 )/2, k2) by adding more
84
sampling points. After the sampling points, k q , for the test function are determined,
the computation of the nodal displacements {U} is repeated for these kq and the
displacements at r i are determined by numerically evaluating the integral of equation
(6). The integration is truncated at kq = J( when both 1q'>(kq)1 and the maximum
amplitudes of the displacements {U} over the canyon surface are smaller than a
prescribed tolerance. Numerical results show that this adaptive scheme associated
with the test function of equation (49) works very well for a variety of systems and
frequencies.
Truncation and Discretization Errors
As mentioned earlier, the infinite boundary integral on the half-space surface in
equation (13) is replaced by an integral over a finite range, L y (equation 22). Thus,
the accuracy in the computation of matrix [A(Xi,kq )] summarized in step 4(b) of the
procedure summary is directly related to L y • It is found that the L y necessary to
achieve a desired accuracy mainly depends on the wavenumber k and the excita
tion frequency w, and is related to the decay of the displacements {u(xo,k)} along
f\. Therefore, the variations of displacements on the half-space surface for various
wavenumber k q is examined. For this purpose, a rigid foundation of width b = 0.2L
on a semi-circular canyon of radius L (Figure 6a) is considered. The foundation in
terface is discretized evenly into 12 four node elements along circumference. Since,
as mentioned in step 7 of the procedure summary, the displacements are expressed
III terms of the nodal tractions {T}, it is only necessary to show the variation of
displacements associated with some of the nodal tractions. The amplitudes of the
Fourier transforms of displacements along the canyon and the half-space boundaries,
f'c U f h , due to horizontal tractions associated with the first node on ri (Figure 6b)
are plotted for several wavenumbers in Figures 7 and 8. These displacem~nts are
obtained from the second column of the matrix [A(kq)]-l[B(kq)] (equation 31), i.e. by
specifying {T} = {O, 1,0, ... , O}T in the equation. The normalized frequencies for the
harmonic tractions are ao = wL/cs = 1 and 2 with corresponding Rayleigh wavenumbers
85
node 1L
x
Ly
z
f\
y
Fig.6a Dam-foundation interface on a semi-circular canyon.
y
Fig.6b Horizontal tractions associated with node 1.
86
0.000-10 -8 -6 -4
I I
-2 a 2Distance y/L
4\
6 8 10
Fig.7 Variation of the amplitudes of Fourier transforms of displacementswith distance; a o = 1.0, v = 1/3. Displacements are due to horizontal tractions associated with the first node for various values ofwavenumber, k.
87
0.000-10 -8 -6 -4 -2 0 2
Distance y/L4 6 8 10
Fig.8 Variation of the amplitudes of Fourier transforms of displacementswith distance; a o = 2.0, v = 1/3. Displacements are due to horizontal tractions associated with the first node for various values ofwavenumber, k.
88
kr =1.07 and kr =2.15.
The displacements decay slowly with distance for wavenumber k q close to the
Rayleigh wavenumber k r , not as slowly for k q < k r +0.5, and very fast for k q > k r +0.5
(Figures 7 and 8). An explanation for the dependence of the decay rate of {u(x,kq )}
with distance on k q can be obtained by examining the asymptotic expressions for the
transformed Green's functions where the decaying trend is dominated by the modified
Bessel functions /(o(asd) and /(1(asd) with as = Jk~ - k; and d being the distance
between source and receiver points (Appendix C of Part I). It should be noted that kr
is larger than, but very close to k s • Thus, when kq > kr +0 where 0 is a small positive
number depending on the size of the system, as is a complex number with positive
real part which increases with kq • For large arguments, the asymptotic expressions
for /(o(asd) and /(1(a sd) are given by
(52a)
(52b)
respectively. As a result, /(0 and /(1 decay exponentially with respect to d which
leads to rapid decay of {u(xo , k q )} along r h • When k q ~ kr + 0, the real part of as
is still positive, but it is very small ( for zero damping, e.g., Re(as) = 0 for ks < ks)
which makes the decay much slower.
Comparing Figures 7 and 8, it can be observed that the displacements decay
more rapidly along the half-space surface for high frequency excitation. Therefore,
for computational efficiency, the discretization range L y should vary with wavenumber
kq and frequency w.
For the rigid banded foundation shown in Figure 6, the impedance matrix with
respect to the mid-bottom point 0' of the foundation can be obtained from equation
89
(46) and expressed as
(53)
o LSxm 0LSyr 0 0000
L2Srr 0 0o L 2 S mm 0o 0 L 2 S tt
Sxx, Syy and Szz are the translational
Sxx 0 0o Syy 0o 0 Szz
o LSyr 0LSxm 0 0
o 0 0
where J.L is the shear modulus for the half-space,
impedance coefficients, Srr and Smm are the rocking impedance coefficients, Stt is
the torsional impedance coefficient, and Syr and Sxm are the coupling terms. All the
coefficients are in a non-dimensional form. The variation of impedance coefficients with
Ly is shown in Figures 9 to 12 where the smooth curves are obtained by connecting
the results at L y / >'8= 0.0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5 and 4.0 through a cubic
spline approximation and the impedance coefficients are normalized by their "exact"
values which are obtained by using L y = 5>'8' In this example, different L y are selected
for k q 5 kr + 0.5 and k q > kr +0.5, and the L y appearing in the figures refers to the
discretization range used for k q :$ k r + 0.5. The relative errors decrease faster with
respect to L y for higher frequency and, for most coefficients, they become less than
two percent if Ly > 2>'8' For a o = 2, the relative errors for the imaginary parts of
the vertical impedance function Szz and the coupling term Syr are relatively large,
but they are within ±4%. Thus, for k q 5 k r + 6, L y should be at least 2>'8' Since
the shear wavelength for low frequency is larger than that for high frequency, larger
discretization range is required for low frequency in general. However, because k r
decreases with frequency and coarse discretization is adequate for low frequency, the
total computational cost due to large Ly for low frequency case does not necessarily
Increase.
For k q > k r +6, a shorter discretization range L y is adequate because of the rapid
exponential decay of the transformed Green's functions. In this investigation, L y for
kq > kr +6 is chosen according to
(54)
90
..........cJ 1.0.........
Q)
a::
0.9
1. 1--
~-- -~ 10 ---- ~- ....... -en . , -~-., -- ""---.........
Q) -a:: --
0.9 I I I I I I I
0 1 2 3 4
ly/As
Fig.9 Real parts of the impedance coefficients determined with various valuesof L y , defining the discretization range on the half-space surface. Theimpedance coefficients are normalized by their "exact" values. Resultsare presented for ao=2 and 4.
91
---
..-Vl~ 1.0'-'
Q)
a::
0.9
1.1
..-evi 1.0'-' -
Q)
a::
0.9o
I
------
I1
I I3
I
4
Fig.IO Real parts of the impedance coefficients determined with various valuesof L Y1 defining the discretization range on the half-space surface. Theimpedance coefficients are normalized by their "exact" values. Resultsare presented for ao=2 and 4.
92
-vr 1.0'-"
E
1.10 0 = 2
- - - - - - - - 0 0 = 4
0.9
1.1
-(/)~ 1.0'-"E
0.9
1.1
-elf 1.0'-"
E
0.9
1.1
-uf 1.0'-"
E
0.9
Fig.ll Imaginary parts of the impedance coefficients determined with variousvalues of L}/, defining the discretization range on the half-space surface.The impedance coefficients are normalized by their "exact" values.Results are presented for ao =2 and 4.
93
1.1
0.9
1.1
.......eJ 1.0
..-E
- 2= 4
0.9
1.1
.......(/)~ 1.0..-E
0.9
1.1
.......J 1.0..-E
0.94
~ 1.0..-E
Fig.12 Imaginary parts of the impedance coefficients determined with variousvalues of L]I' defining the discretization range on the half-space surface.The impedance coefficients are normalized by their "exact" values.Results are presented for ao=2 and 4.
94
which decreases with increasing k q • To ensure the accuracy of the results on fe,
however, L y should not decrease without limit. Since the errors in the computed
displacements near the end of the discretization on l'h are relatively large, it is
recommended that L y should be greater than the smaller one of As and the half
width of the canyon.
The size of elements on the half-space boundary f' h may vary depending on their
distances from the canyon. In this study, element sizes on l'h are chosen according to
an arithmetic series (I, 1+ ~I, 1+ 2~1, ... ) where the elements adjacent to the canyon
have the same size as their neighbors on f' c and the elements farthest away have
the size As /4. Furthermore, in order to avoid ill-conditioning or singularity in the
flexibility influence matrix [f] (equation 39), there are some restrictions on location of
nodes on the canyon boundary f' c relative to the locations of nodes on r i. Detailed
description and theoretical background for this restriction are given in Appendix C.
In summary, the following guidelines for selection of kq , J(, L y and the element
sizes are recommended:
1. The sampling points, k q , in the wavenumber domain should be chosen unevenly
with finely-spaced sampling points near the Rayleigh wavenumber and coarsely-spaced
sampling points elsewhere. Either piecewise Gauss integration scheme or an adaptive
integration procedure can be used to determine k q • The adaptive procedure may be
combined with a test function which characterizes the main features of the Fourier
transforms of the displacements and the integral in equation (6) may be truncated
when the magnitudes of the test function and the nodal displacements {iT} on canyon
boundary are smaller than a prescribed tolerance.
2. The choice of the integration range, L y , on the half-space surface should depend
on k q • For k q ~ k r + 6, where 6 is a small positive number depending on the' size of
the system, L y > 2As where As is the shear wavelength. For kq > kr +6, L y may be
chosen according to equation (54) with the restriction that L y is greater than the
95
smaller of two values: As and the half width of' the canyon.
3. The element size on both the dam-foundation rock interface, r i, and the
boundary f e should be kept smaller than As /5.
VERIFICATION AND COMPUTATIONAL COSTS
A computer program was developed to implement the step by step procedure
summarized earlier to determine the impedance matrix of the foundation rock region
shown in Figure 1 with reference to the degrees of freedom on the dam-foundation
rock interface rio Results for two systems are presented to demonstrate verification
of the present procedure.
The first system analyzed is a rigid square footing on the surface of a half-space for
which the impedances or compliances are available [10,18]. Figure 13 shows some of
the compliance functions determined by various methods for different non-dimensional
frequency ao = wb/cs where b is the half-width of the footing. The hysteretic damping
factors used in the present analysis are TIp = TIs = 0.001. In applying the present method,
the interface between the footing and the half-space is discretized into thirty six 8
node, square elements, the line segment (-b,b), which is equivalent to fe, is discretized
into 12 elements, and the boundary fh on the half-space surface is discretized up
to L y = 2As for kq ~ k r +0.5 where As is the corresponding shear wavelength. The
agreement between various results shown in Figure 13 is satisfactory. The differences
may be, in part, due to the finite discretization range, L y , on the half-space surface,
which leads to a stiffer system in the present method, and the slight difference in
damping used in the various analyses.
The second system analyzed is an infinitely-long canyon of semi-circular cross
section of radius L with a dam-foundation rock interface of width b = 0.2L (Figure
14). The impedance functions are determined for the interface ri assumed to be rigid
with six degrees of freedom. In implementing the present procedure, the interface is
discretized evenly into 12 elements along the circumferential direction and 2 elements
96
0.20
0.15
Wong & Luco [1 0]- - - - - - - - - Rizzo et.al. (18]
o Present Results
.01 0.10u
0.05
0.000.2
-----
41 2 3Normalized Frequency, 0 0
o0.0
.0
=!- 0.1J:
u
0.00.3
Real0.2..,
.0::t22u 0.1
Fig.13 Comparison of compliances obtained by present method with previousresults for vertical (Cvv), horizontal (CH H) and rocking (CM M )compliances of square footing supported on the surface of a viscoelastichalf-space (v = 1/3).
97
(a) Dam-foundation interface
Lx
O.1L
x
~~~~C L Ly y
(b) Plan view of a quarter of an illustrative (not actual) three-dimensionalboundary element discretization
Fig.14 Dam-foundation interface on a semi-circular canyon of radius L.
98
along the canyon axIS. The boundary l'c is discretized evenly into 12 line elements
and L y = 2As for kq < kr + 0.5. The impedances are presented in Figures 15 and
16 as a function of the normalized frequency ao = wL/cs . These impedance func
tions result from cubic spline functions connecting computed impedance values for
ao = 0.2,0.5, 1.0, 1.5,2.0,2.5,3.0,3.5 and 4.0. Because previous results are apparently not
available for this system, the impedances were also obtained by analyzing the same
system by a three-dimensional boundary element method (equation 8). The interface
r j , a finite length Lx = L of the canyon, and the half-space surface up to L y = L
are discretized by two-dimensional boundary elements (Figure 14). The agreement is
satisfactory between the results obtained by the 3-D boundary element method and
by the procedure presented in this report.
Compared with the general 3-D boundary element approach, the present method
IS more accurate since the boundary integration in the direction of canyon axis IS
evaluated analytically. If highly-accurate results are desired, the present method IS
also more efficient. This is illustrated in Table 1 where the relative errors in the
impedance coefficients for the second problem described above are listed. The system
was analyzed by the 3-D boundary element method with L y = L and various values
of Lx, and by the present method with L y = L. Lacking an exact solution for this
problem, the comparison is performed against the solutions by the present method with
the discretization range L y = 2As • Clearly, the performance of the present method is
superior to that of the 3-D boundary element methods in accuracy and computational
time if results with errors less than 5% are desired.
The present method is especially efficient when the number of DOF along the
canyon axis is relatively large (Appendix B). Since the computer program implementing
the 3-D boundary method was not written with blockwise storage and, hence, is limited
by the computer storage capacity, relatively short discretization ranges L y and Lx
were used (Table 1). In general, however, such short discretization range are not
expected to lead to accurate results because the results are likely to oscillate with L y
99
6~--------------,
Real
_o-----o-----~-------4
6-r---------------,5Real
,.. -----9 -----e----- ------
5
4
xx 3
(/)~3
(/)
2 ...... Imag./ao--0------0------0- _
2 Imag./ao- - E>- _~-----G--- __
o 2 3 4 o 1 2 3 4
(a) Translational Impedance So (b) Translational Impedance S»,
6-r---------------. 6--r---------------,
5 Real 5 Real
4D
_--..0. _--_--0----- 4 -----o-----O- ~----
Imag./ao--El_
- .. --'O"-----G-- _
432
Imag./co
_ -G-----e-----~----
o
2
.. 3en
4321o
2
NN 3
(/)
(c) Translational Impedance Su (d) Rocking Impedance S"
Fig.15 Comparison of foundation impedances obtained by the present method(dashed lines) with the 3-D boundary element method with Lx = L(symbols). The system analyzed is a rigid foundation of width 0.2Lon a semi-circular canyon of radius L (v = 1/3, Tip = Tis = 0.02).
100
4~-----------'----, 4-,.----------------,
.3 .3 Real
E 2E
(f) Real-----~-----o-----e-----
Imag.jao',---G-----e-----~----
:=2(J)
... "' .0. - - - - -, - -0 - - - - - ..,
Imag./ao
_-6 - - - - - 0_ - - - - -0- - - --
o 2 .3 4 o 2 3 4
(a) Rocking Impedance Smm (b) Torsional Impedance Sit
4--.----------------,
_---o-----~----_ -G-
4
Imag.jao
2 3
Real
- - 9- - - - - -& - - - _ - - -0 - --
'--- G- -----0-'-- '"-1J"- _
o
o
-2
E -1"(J)
432
Real3~~2~
(J) ,
--1 Imag./ao1 ~ ..1----0 - - - - - 0- - - - - -<>- - - - -
o -L'...,.....,.-r-~-,--r-r-,-r--r--,--r_T""...,..__.....,.........,,..._1I I ,
o
(c) Coupling Impedance S1" (d) Coupling Impedance S"",
Fig.16 Comparison of foundation impedances obtained by the present method(dashed lines) with the 3-D boundary element method with Lx = L(symbols). The system analyzed is a rigid foundation of width O.2Lon a semi-circular canyon of radius L (v = 1/3, TJp = 7]8 = 0.02).
101
Table 1 Percentage Errors in Impedances and Computation Time for Two Methods
3-D Boundary Element Method with L", = Present MethodImpedances
0.0 0.25L 0.5L 0.75L 1.0L with L", = L
Re(S",,,,) 7.0 4.2 4.1 4.1 3.7 0.6
Im(S",,,,) 18.4 10.5 7.3 4.7 2.1 -1.3
Re(Syy) -7.9 3.6 4.0 2.6 1.3 0.2
Im(Syy) 15.3 8.7 1.6 -1.0 -1.5 2.6
Re(Szz) 1.8 3.2 2.2 1.9 1.9 2.0
Im(Szz) 10.4 6.3 5.0 5.3 5.4 -3.8
Re(Srr) -5.5 -1.1 1.1 2.6 2.9 0.8
ao = 2 Im(Srr) 7.8 8.2 6.9 4.5 1.1 3.0
Re(Smm) 13.9 7.8 4.2 2.7 2.1 0.5
Im(Smm) -3.1 -8.4 -6.3 -3.7 -2.2 -1.0
Re(Stt) -3.5 -6.6 -4.9 -1.8 0.8 0.0
Im(Stt) -9.8 -1.7 5.7 9.7 8.3 1.2
Re(Syr) -1.2 1.3 2.6 2.3 1.3 0.8
Im(Syr) 13.6 7.6 2.3 -1.1 -3.4 2.7
Re(S",m) 14.7 9.8 6.9 5.5 4.4 0.8
Im(S",m) 13.2 3.2 1.1 -0.02 -1.6 -1.7
Re(S",,,,) 0.02 -2.9 2.1 3.2 0.3 0.5
1m(S",,,,) -4.1 3.8 5.9 2.3 0.8 -0.1
Re(Syy) 4.8 4.2 2.8 3.8 4.3 0.2
Im(Syy) 4.4 -1.3 0.3 1.5 -0.4 1.3
Re(Szz) -1.3 -1.3 1.8 2.4 -0.8 1.0
Im(Szz) -1.3 2.3 2.9 0.6 0.04 0.1
Re(Srr) 9.9 3.2 0.5 1.5 5.1 -1.9
ao = 4 Im(Srr) 2.3 -3.6 -1.2 1.7 -0.8 1.7
Re(Smm) -9.4 -4.8 1.5 1.6 0.0 -0.1
Im(Smm) -0.6 7.6 7.6 4.1 4.5 -0.03
Re(Stt) 18.7 5.8 3.6 3.8 2.6 -0.3
Im(Stt) 3.6 -1.7 0.2 0.7 -2.0 0.6
Re(Syr) 14.5 2.0 0.9 2.3 3.8 -1.7
Im(Syr) 0.1 -0.8 1.6 3.8 0.7 2.8
Re(S",m) -8.0 -6.5 0.2 1.9 -0.3 0.5
Im(S",m) -7.2 4.2 7.0 4.0 3.5 -0.1
Compt.Time* (sec.) 3.1 7.9 15.6 26.8 41.4 24.0
* on Cray X-MP/14
102
and Lx (Table 1) until both discretization ranges extend two times the shear wave
length.
The present method has the advantage that it requires much less storage compared
to the 3-D boundary method. Thus the accuracy of the present method can be
improved at extra computational cost by increasing the discretization range and
making the discretization finer. However, the accuracy of the 3-D boundary method
is likely to be limited by the available computer storage.
CONCLUSIONS
A direct boundary element procedure is presented to determine the impedance
matrix, i.e. the frequency-dependent stiffness matrix, associated with the nodal points
at the base of a structure supported on a canyon cut in a homogeneous viscoelastic
half-space. The canyon is infinitely-long and may be of arbitrary but uniform cross
section. The uniform cross-section of the canyon permits analytical integration along
the canyon axis of the three-dimensional boundary integral equation. Thus the original
three-dimensional problem is reduced to an infinite series of two-dimensional boundary
problems, each of which corresponds to a particular wavenumber and involves Fourier
transforms of full-space Green's functions. Appropriate superposition of the solutions
of these two-dimensional boundary problems leads to a dynamic flexibility influence
matrix which is inverted to determine the impedance matrix.
A step-by-step summary of the procedure is presented and numerical aspects of
the method have been investigated. In particular, guidelines are presented for selection
of the sampling points in the wavenumber domain, the finite discretization range on
the half-space surface necessary to approximate the infinite range, and the element
size on the dam-foundation rock interface.
The accuracy of the procedure and implementing computer program has been
verified by comparison with previous results for a surface-supported, square foundation
and solutions for a foundation of finite-width on a infinitely-long canyon by a three-
103
dimensional boundary element method (BEM). Compared with the three-dimensional
BEM, the present method requires less computer storage and is more accurate and
efficient. Computation of the foundation impedance matrix by this method would
enable earthquake analysis of arch dams including dam-foundation rock interaction
effects:
104
REFERENCES·
1. G. Fenves and A.K. Chopra, "Earthquake Analysis of Concrete Gravity DamsIncluding Reservoir Bottom Absorption and Dam-Water-Foundation Rock Interactions," Earthquake Engineering and Structural Dynamics, 12, pp.663-680, 1984.
2. K.-L. Fok and A.K. Chopra, "Earthquake Analysis of Arch Dams Including DamWater Interaction, Reservoir Boundary Absorption and Foundation Flexibility,"Earthquake Engineering and Structural Dynamics, 14, pp.155-184, 1986.
3. P.S. Nowak, "Effect of Nonuniform Seismic In:2.ut on Arch Dams," Report No.EERL 88-03, California Institute of Technology, Pasadena, California, 1988.
4. Y. Ghanaat and R.W. Clough, "EADAP Enhanced Arch Dam Analysis Program,User's Manual," Report No. UCB/EERC-89/07, Earthquake Engineering ResearchCenter, University of California, Berkeley, Nov. 1989.
5. J.P. Wolf and G.R. Darbre, "Dynamic-stiffness Matrix of Soil by The BoundaryElement Method: Embedded Foundation," Earthquake Engineering and StructuralDynamics, 12, pp.401-416, 1984.
6. E. Alarcon, J. Dominguez and F.del Cano, "Dynamic Stiffness of Foundations,"New Developments in Boundary Element Methods, Proc. 2nd into seminar recent advancesboundary element methods, (ed. C.A. Brebbia) , pp.264-280, 1980.
7. T-J. Tzong and J. Penzien, "Hybrid Modeling of a Single-Layer Half-Space Systemin Soil-Structure Interaction," Earthquake Engineering and Structural Dynamics, 14,pp.517-530, 1986.
8. R.J. Apsel, "Dynamic Green's Functions for Layered Media And Applications toBoundary-value Problems," Ph.D. Dissertation, University of Califorma, San Diego,CA, 1979.
9. R.J. Apsel and J .E. Luco, "Impedance Functions for Foundations Embedded in aLayered Medium," Earthquake Engineering and Structural Dynamics, 15, pp.213-231,1987.
10. H.L. Wong and J .E. Luco, "Impedance Functions for Rectangular Foundations ona Uniform Viscoelastic Half-space," J. Geotechnical Engineering, ASCE, submittedfor publication.
11. J .E. Luco and H.L. Wong, "Response of Hemispherical Foundation Embeddedin Half-Space," J. Engineering Mechanics, ASCE, 112, No. 12, pp.1363-1374, Dec.1986.
12. H.L. Wong and J .E. Luco, "Tables of Impedance Functions for Square Foundationson Layerea Media," Soil Dynamics and Earthquake Engineering, 4, No.2, pp.64-81,1985.
13. C-H. Chen and J. Penzien, "Dynamic Modelling of Axisymmetric Foundations,"Earthquake Engineering and Structural Dynamics, 14), pp.823-840, 1986.
14. J. Dominguez, "Dynamic Stiffness of Rectangular Foundations," Research ReportR78-20, Dept. of Civil Engineering, MIT, Cambridge, MA, 1978.
15. J .M. Emperador and J. Dominguez, " Dynamic Response of Axisymmetric embedded Foundations," Earthquake Engineering and Structural Dynamics, 18, pp.1l05-1l17,1989.
16. A.P. Gaitanaros and D.L. Karabalis, "Dynamic Analysis of 3-D Flexible Embedded Foundations by a Frequency Domain BEM-FEM," Earthquake Engineering andStructural Dynamics, 16, pp.653-674, 1988.
105
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.34.
F.J. Rizzo, D.J. Shi2PY and M. Rezayat, "A Boundary Integr_al Equation Methodfor Time-harmonic Radiation and Scattering in an Elastic -Half-space," AdvancedTopics in Boundary Element Analysis, ASME, AMD, 72, pp.83-90, 1985.F.J. Rizzo, D.J. Shippy and M. Rezayat, "Boundary Integral Equation Analysisfor a Class of Earth-Structure Interaction Problems," Final Report, Dept. Eng.Mechanics, Univ. of Kentucky, Lexington, Kentucky, 1985.D.L. Karabalis and D.E. Beskos, "Dynamic Response of 3-D Embedded Foundations by The Boundary Element Method," Computer Methods in Applied Mechanicsand Engineering, 56, pp.91-119, 1986.M.J. Kaldjian, "Torsional Stiffness of Embedded Footings," Soil Mech. Found. Div.,ASCE, 95, No.SM1, pp.969-980, 1969.C.M. Urlich and R.L. Kuhlemeyer, "Coupled Rocking and Lateral Vibrations ofEmbedded Footings," Can. Geotech., 10, 2, pp.145-160, 1973.E. Kausel, J.M. Roesset and G. Waas, "Dynamic Analysis of Footings on LayeredMedia," J. Engineering Mechanics, ASCE, 101, No.EM5, pp.679-693, 1975.E. Kausel and J.M. Roesset, "DJnamic Stiffness of Circular Foundations," J.Engineering Mechanics, ASCE, 101, No.EM6, pp.771-785, 1975.S.M. DaY7,.T"Finite Element Analysis of Seismic Scattering Problems," Ph.D. Dissertation, university of California, San Diego, CA, 1977.V.W-S. Lee, "Investigation of Three-Dimensional Soil-Structure Interaction," Report CE 79-11, Dept. of Civil Engineering, University of Southern California, LosAngeles, CA, 1979.Y.K. Chow and I.M. Smith, "Infinite Elements for Dynamic Foundation Analysis,"Numerical Methods in Geomechanics, (Ed. Z. Eisenstein), 1, pp.15-22, 1982.
F. Medina and J. Penzien, "Infinite Elements for Elastodynamics," EarthquakeEngineering and Structural Dynamics, 10, pp.699-709, 1982.S. Gupta J. Penzien, T.W. Lin and C.S. Yeh, "Three-Dimensional Hybrid Modelling of Soil-Structure Interaction," Earthquake Engineering and Structural Dynamics,10, pp.69-87, 1982.A. Mita and W. Takanashi, "Dynamic Soil-Structure Interaction Analysis byHybrid Method," Boundary Elements, C.A. Brebbia, T. Futagami and M. Tanaka,Eas., pp.785-794, 1983.J.E. Luco, "Linear Soil-structure Interaction: A Review," Earthquake Ground MotionAnd Its Effects on Structures, AMD-Vol. 53, S.K. Datta (Ed.). American Societyof Mechanical Engineers, New York, pp.41-57, 1982.D.E. Beskos, "Boundary Element Methods in Dynamic Analysis," Applied MechanicsReviews, 40, 1, pp.1-23, 1987.F. Hartmann, "Computing the C-matrix in non-smooth boundary points," New Developments in Boundary Element Methods, (Ed. by C.A.Brebbia), CML Publications,pp.367-379, 1980.K.F. Graff, Wave Motion in Elastic Solids, Ohio State University Press, 1975.Pei-cheng Xu and A.K. Mal, "An Adaptive Integration Scheme for IrregularlyOscillatory Functions," Wave Motion, 7, pp.235-243, 1985.
106
[A]
b
[B)
[13]
c
[C(X)]
[I]
[1]
NOTATION.
non-dimensional frequency defined by ao = wL/c.
= Jp -k;=Jk2
- k;square coefficient matrix for the boundary element system (equation29)
3 x 3Mn matrix assembled from the integrals of the products of[Ff and [N] associated with a single source point (equation 25)width of a foundation or half-width of a footing
matrix assembled from [B] (equation 30)
3 x 3Mi matrix assembled from the integrals of the products of [G]Tand [R] associated with a single source point (equation 26)matrices for the decomposed expression of matrix [B]
=w/kP wave velocity including damping
P wave velocity without damping
Rayleigh wave velocity including damping
Rayleigh wave velocity without damping
shear wave velocity including damping
shear wave velocity without damping
3 x 3 matrix related to the geometric shape of the boundary at x(equation 14)
= J(y - y.)2 + (z - z.)2, distance between source and receiver pointsprojected on plane x = 0
3 x 3 diagonal matrix defined by [b] = diag(l, -1, -1)
diagonal matrices consisting of [b) (Appendix B)
3 x 3Mi ma.trix relating the Fourier transform of the displacementsto the nodal tractions at r i (equation 35)flexibility influence matrix
3 x 3Mi matrix through which nodal displacements on r i are expressed in terms of {T}
107
[F]
[F-]
g(k)
[G]
[G-]
rhO]
rhO]
k
K
I
L
M
M
transformed Green's traction functions (equation llb)
full-space Green's traction functions
= (2Jc2 - k;)2 - 4k2(k2 - k;)l/2(Jc2 - k;)l/2, Rayleigh function
transformed Green's displacement functions (equation lla)
full-space Green's displacement functions
matrix of interpolation functions for two-dimensional elements on
r.Fourier transform of the two-dimensional interpolation functions
rhO]matrices enlarged from matrices {~} and ['11]-1 (Appendix A)
matrix assembled from [hOl
=Aintegration of the test function t/>(k) over an interval
wavenumber in x-direction
qth sampling point in the k wavenumber domain
Rayleigh wavenumber
shear wavenumber
values in the wavenumber domain
integration limit in the k wavenumber domain at which the infiniteintegral of equation (6) is truncatedmodified Bessel functions of second kind of order zero and orderone
length of an element on f'c U f'h
half width of a two dimensional canyon
discretization range along the direction of canyon axIS for threedimensional boundary element approachdiscretizat on range on the half-space surface
number of one-dimensional elements on the boundaries f c U f'h
number of nodes on an element
number of Gauss integration points on l'c
number of nodes on the dam-foundation rock interface r i
108
{P} l P", , Py , P",
{r}
{il}
[8]
number of nodes on the boundaries f c U f h
element number for the first and the last one-dimensional elementson boundaryfc
outgoing normal to the boundary and its components
shape function matrix for the one-dimensional element on f c ufh ;
jth shape functionvector and components of concentrated forces associated with thesecond elastodynamic statevector of nodal displacements at the dam-foundation rock interface
6 x 1 vector containing the six degrees of freedom for a rigid foundationradius of the circular contour around a source point in plane x = 0
vector of nodal forces at the dam-foundation rock interface r i
unsymmetric impedance matrix resulted from boundary elementapproximationsimpedance matrix
impedance matrix for a rigid foundation
Stt, Smm, S",m, Syr non-dimensional impedance coefficients for rigid foundation
t
{t};t""ty,tz
{l};f"" fy, t",
{T}
{T",}, {Ty }, {T",}
{u}
{iL}
x,y,z
time
traction vector and its components
vector of Fourier transform of the surface tractions and its componentsvector of nodal tractions
vector containing the x-, y-, and z-components of nodal tractions
displacement vector
Fourier transform of the displacement vector {u}
vector containing nodal displacements at plane x = 0 and its components at the jth nodeweighting coefficients associated with the integration scheme usedin k wavenumber domainspatial coordinates
109
x',y',z'
x" y" z,
[aJ
'1r
.A,
1I
{¢}
spatial coordinates in x'y'z'-coordinate system
coordinates of a point inside an element on r i
spatial coordinates of source point
= (x, y, z), an arbitrary point
a point in x'y'z'-coordinate system
= (0, y, z), an arbitrary point at plane x = 0
= (0, y" z,), source point at plane x = 0
3M. x 6 matrix relating nodal displacements on r i with the SIX
rigid-body degrees of freedom of a rigid foundationcanyon surface
cross-section of the canyon at plane x = 0
half-space surface
cross-section of the half-space surface at plane x = 0
dam-foundation rock interface
extent of the jth element
small contour of radius r, around the source point at plane x = 0
positive number
Kronecker delta function
natural coordinate
constant hysteretic damping factor for P wave
constant hysteretic damping factor for Rayleigh wave
constant hysteretic damping factor for shear wave
shear wavelength
shear modulus for the half-space medium
Poisson's ratio for the half-space IT_edium
test function defined in equation (49)
matrix that transforms distributing tractions into concentratednodal forces (equation 41)
row matrix defined in equation (A.3)
110
{,,&}j "&i[\l1]
w
Fourier transform of {1/J} and its jth component
square matrix defined in equation (A.4)
frequency in rad/sec.
111
APPENDIX A: FOURIER TRANSFORMS OF TWO-DIMENSIONAL
SHAPE FUNCTIONS
First, consider a two-dimensional curved 8 node element at the dam-foundation
rock interface fi (Figure A.l) where any point, X, on the element can be uniquely
represented by two local coordinates x' = (x', yl) (or (x', Zl) if the element is on a
vertical plane). The local and the global coordinate systems are related through
x = Xc +x'
y = Yc + y'
where (xc, Yc) are the global coordinates of a point inside the element.
(A.la)
(A.lb)
As described in the main text, the prescribed traction, {t(x)}, contains the x, y
and z components. For the purpose of illustration, consider the X component, tx(x),
first. At any point x' on the element, the traction tx(x' ) can be expressed in terms
of tractions at all the nodal points, x~, i=I,... ,8:
(A.2)
(A.3)
(AA)
where
({¢(~d})
[\II] = .
{¢(xs)}
and {Tx} = {Txt, ... , Txs}T is a vector containing the x-components of the nodal tractions
t --I '-1 8a Xi' Z- , ••• , •
Utilizing equations (A.l) and (A.2), the Fourier transform of tx(x) can be found
to be
112
x
x'(~,y)
IIIII
:(~,Yc y'I
(X01
,Y)II
o y
Fig.A.l Top VIew of an arbitrary element on dam-foundation interface.
113
(A.5)
where X o1 and X o2 are the global x-coordinates of the points (Figure A.I) at the element
boundary intersected by a line parallel to the x-axis passing through x' = (0, v'). In
equation (A.5), {~} is a 1 x 8 matrix containing the Fourier transforms of {'I/7(i')}
where
- 1 i Ok ok'1/71 = (e l X 0 2 _ el X 0 1 )
21r k
.1. 1[ i( ) 1] ikx 1[i ( ) 1] ikx0/2 = 211' -k X o2 - Xc + k2 e 02 + 21r k X o1 - Xc - k2 e 01
~3 = ~lY'
- 1 ikx [1 2 2 2i]'1/74 = 21r e 02 ik (X o2 - xc) + k 2 (X o2 - xc) + k3
1 ikx 0 1 [ 1 ( )2 2 ( ) 2i]- 211' e ik Xol - Xc + k2 Xol - Xc + k3
~5 = ~2Y'
~6 = ~ly'2
~7 = ~4Y'
~8 = ~2y'2
(A.6)
Similarly, the Fourier transforms of the prescribed tractions in the y and z direc
tions can be expressed as
114
(A.7)
(A.8)
where {Ty} and {Tz } are the y and z components of the nodal tractions, respectively.
Combining equations (A.5), (A.7) and (A.8) leads to the expression for the 3 x 24
matrix [h(xo,k)] (equation 19) for the 8 node element
(A.9)
where [hI] is a 3 x 24 matrix enlarged from matrix {~} by replacing each entry in
{~} by a 3 x 3 diagonal matrix diag[1/Ji' ~i, 1/Ji], i = 1, ... ,8; and [h2 ] is a 24 x 24 matrix
enlarged from matrix ["iI!]-I in a similar manner.
For 4 and 6 node elements, the solution procedure is the same except that {¢(x')}
in equation (A.3) is replaced by {'ljJ(i')} = {l,x',y',x'y'} for 4 node element, and by
{'ljJ(x')} = {1,x',y',x12 ,x'y',y'2} for 6 node element.
115
APPENDIX B: RELATION BETWEEN {Uh AND {U}-k
As mentioned in the procedure summary, the computations to determine the
nodal displacements {U} are repeated for a discrete set of wavenumbers over a trun
cated wavenumber range (-K, K). Due to a special property of the integrands with
respect to k, however, {U} for negative wavenumbers can be obtained indirectly from
{U} corresponding to positive wavenumbers. From Appendix C in Part I, the trans-n
formed Green's functions [G] and [F] corresponding to negative wavenumber -k can
be expressed as
where [D] is a 3 x 3 diagonal matrix given by
[D] = diag(l, -1, -1)
(B.1a)
(B.1b)
(B.2)
Thus, the coefficient matrix [A(k)] in equation (28), which consists of the in-n
tegrals of the products of the transformed traction Green's functions [F] and the
one-dimensional interpolation functions, has the property that
[A( -k)] = [D][A(k)][D]
where [D] is a diagonal matrix of order 3Mn
[D] = diag(l, -1, -1, 1, -1, -1, ..., 1, -1, -1)
From equation (B.3), it is apparent that
[A( _k)]-l = [D][A(k)r1[D]
since [D]-l = [D].
(B.3)
(BA)
(B.5)
The matrix [B(k)] in equation (28) consists of the integrals of the products of [G],
the transformed displacement Green's functions, and [H(io , k)], the Fourier transforms
116
of the two-dimensional interpolation functions on the dam-foundation rock interface r i
(equations 26 and 30). These integrals are evaluated numerically over each elements
on i:'c through Gauss integration scheme. Since [H(xo,k)] does not depend on the
source point xos , matrix [B(k)] can be decomposed as
[B(k)] = [Ba(k)][BH(k)] (B.6)
where matrix [Ba(k)] consists of the transformed displacement Green's functions asso
ciated with different source points xos and receiver points xo , i.e. the Gauss integration
points; and matrix [BH (k)] consists of [H(x0' k)] for various receiver points. The sizes
of [Ba(k)] and [BH(k)] depend on the total number of Gauss integration points on
f'c' If the total number of Gauss integration points on f'c is Mg , the sizes for [Ba(k)]
and [BH(k)] are 3Mn x 3Mg and 3Mg x 3Mi , respectively. Due to the relation given
in equation (RIa) and the fact that [H(xo, -k)] is complex conjugate of [H(xo,k)],
matrix [B(k)] apparently has the property that
[B( -k)] = [D][Ba(k)][D1]conjg([BH(k))) (B.7)
where [D1] is a square diagonal matrix of dimension 3Mg having similar form as [D]
in equation (BA).
Thus, from equation (31), the nodal displacements on the boundary :rc u :rhare
given by
(B.8)
for positive wavenumber k and
(B.9)
for negative wavenumber -k. Once [A(k)t1[Ba(k)] is found, both {Uh and {U}-k
can be determined easily.
In general, it is more computationally efficient to use equations (B.8) and (B.9),
instead of equation (31), to solve for the nodal displacements {U} on the boundaries
117
f'cUf'h for both k and -k, especially when the total number of nodes on r i is greater
than the total number of Gauss integration points on f'c, i.e. Mi > Mg. Since the
computation of [A(k)]-l[BG(k)] does not depend on the nodal distribution at r i along
the x-direction, the computational cost for the present method does not increase
much when the number of nodes at r i increases along the x-direction. Therefore, the
present method is especially efficient when the number of DOF along the canyon axis
is relatively large.
118
APPENDIX C: RESTRICTIONS ON ·MESH LAYOUT ON f e
Two types of discretization are involved in the present method. One is the
discretization of l'e U f h , the system cross-section at x = 0, into one-dimensional line
elements and the other is the discretization of the dam-foundation rock interface, ri,
into two-dimensional surface elements. In order to avoid singularity problem in the
flexibility matrix, there is a certain restriction on the relative relation between the
two types of discretization.
In order to illustrate the restriction, consider a simple example shown in Figure C.l
where a banded foundation of constant width is resting on a semi-circular canyon. In
this example, l'e is discretized evenly into 4 line elements with the left most node being
the mth node and r i is discretized evenly into 8 elements along the circumferential
direction. From equation (33), the nodal displacements, {r}, on C is determined by
superposing the Fourier transforms {u(ioj,kq)} at fe' where {u(ioj,kq )} is determined
from their nodal values through equation (34). Thus, for this particular example, the
displacements at the first three nodes on ri are given by
{rh = I:Wq[Nl((d{Uq}m + N2((1){Uq}m+t]e-ikqXlq
{r}z = L Wq[N1((2){Uq}m + N2((2){Uq}m+I]e-ikqX2q
{rh = L Wq[NI((3){Uq}m + N2((3){Uq}m+1]e-ikqX3q
(C.l)
(C.2)
(C.3)
where (1, (2 and (3 are the corresponding natural coordinates for the three nodes.
Since Xl = X2 = X3, it is clear that these nodal displacements can be expressed in
matrix form as
(CA)
Because the second matrix on the right hand side of equation (C.4) is a linear
combination of the nodal tractions, {T} (see equation 35), equation (C.4) implies that
119
L
x
y
L
i\01.__---_...£----
y
z
Fig.C.1 lllustration of inappropriate mesh that will lead to singular flexibilityinfluence matrix.
120
the first three rows in the flexibility influence matrix [I] (equations 38 and 39) is a
linear combination of two rows. Thus the rank of the first three rows is less than 3
which means that the resulting flexibility influence matrix is singular.
From this example, it is clear that, in order to avoid singularity or ill-condition
problem in the flexibility influence matrix, the size of the elements on f e should be
smaller than or about the same as that of elements on rio More specifically, situations
similar to the example just illustrated, where three or more nodes having the same
x-coordinates on r i fall in one element on fe, should be avoided.
121
EARTHQUAKE ENGINEERING RESEARCH CENTER REPORT SERIES
EERC reports are available from the National Information Service for Earthquake Engineering(NISEE) and from the National Technical InformationService(NTIS). Numbers in parentheses are Accession Numbers assigned by the National Technical Information Service; these are followed by a price code.Contact NTIS, 5285 Port Royal Road, Springfield Virginia, 22161 for more information. Reports without Accession Numbers were not available from NTISat the time of printing. For a current complete list of EERC reports (from EERC 67-1) and availablity information, please contact University of California,EERC, NISEE, 130 I South 46th Street; Richmond, California 94804.
UCB/EERC-81101 "Control of Seismic Response of Piping Systems and Other Structures by Base Isolation; by Kelly, J.M., January 1981, (PB81 200735)A05.
UCB/EERC-81/02 "OPTNSR· An Interactive Software System for Optimal Design of Statically and Dynamically Loaded Structures with NonlinearResponse; by Bhatti, M.A., Ciampi, V. and Pister, K.S., January 1981, (PB81 218 851)A09.
UCB/EERC-81103 "Analysis of Local Variations in Free Field Seismic Ground Motions; by Chen, J.-C., Lysmer, J. and Seed, H.B., January 1981, (ADA099508)AI3.
UCB/EERC-81/04 "Inelastic Structural Modeling of Braced Offshore Platforms for Seismic Loading; by zayas, V.A., Shing, P.·S.B., Mahin, SA andPopov, E.P., January 1981,lNEL4, (PB82 138 777)A07.
UCB/EERC-81105 "Dynamic Response of Light Equipment in Structures; by Der Kiureghian, A., Sackman, J.L. and Nour-Dmid, B., April 1981, (PB81218497)A04.
UCB/EERC-81/06 "Preliminary Experimental Investigation of a Broad Base Liquid Storage Tank; by Bouwkamp, J.G., Kollegger, J.P. and Stephen, R.M.,May 1981, (PB82 140 385)A03.
UCB/EERC-81/07 "The Seismic Resistant Design of Reinforced Concrete Coupled Structural Walls: by Aktan, A.E. and Benero, V.V., June 1981, (PB82113 358)AIl.
UCB/EERC-81/08 "Unassigned; by Unassigned, 1981.
UCB/EERC-81/09 "Experimental Behavior of a Spatial Piping System with Steel Energy Absorbers Subjected to a Simulated Differential Seismic Input," byStiemer, S.F., Godden, W.G. and Kelly, J.M., July 1981, (PB82 201 898)A04.
UCB/EERC-81/l0 "Evaluation of Seismic Design Provisions for Masonry in the United States; by Sveinsson, B.I., Mayes, R.L. and McNiven, H.D.,August 1981, (PB82 166 075)A08.
UCB/EERC-81/l1 "Two-Dimensional Hybrid Modelling of Soil-Structure Interaction," by Tzong, T.-J., Gupta, S. and Penzien, J., August 1981, (PB82 142118)A04.
UCB/EERC-81/l2 -Studies on Effects of Infills in Seismic Resistant RIC Construction: by Brokken, S. and Bertero, V.V., October 1981, (PB82 166190)A09.
UCB/EERC-81/13 "Linear Models to Predict the Nonlinear Seismic Behavior of a One-Story Steel Frame; by Valdimarsson, H., Shah, A.H. andMcNiven, H.D., September 1981, (PB82 138 793)A07.
UCB/EERC-81/l4 "TLUSH: A Computer Program for the Three-Dimensional Dynamic Analysis of Earth Dams: by Kagawa, T., Mejia, L.H., Seed, H.B.and Lysmer, J., September 1981, (PB82 139 940)A06. .
UCB/EERC-81/15 -Three Dimensional Dynamic Response Analysis of Earth Dams: by Mejia, L.H. and Seed, H.B., September 1981, (PB82 137 274)A 12.
UCB/EERC-81/16 "Experimental Study of Lead and Elastomeric Dampers for Base Isolation Systems: by Kelly, J.M. and Hodder, S.B., October 1981,(PB82 166 182)A05.
UCB/EERC-81/17 "The Influence of Base Isolation on the Seismic Response of Light Secondary Equipment: by Kelly, J.M., April 1981, (PB82 255266)A04.
UCB/EERC-81118 -Studies on Evaluation of Shaking Table Response Analysis Procedures: by Blondet, J. M., November 1981, (PB82 197 278)AI0.
UCB/EERC-81119 -DELIGHT.STRUCT: A Computer-Aided Design Environment for Structural Engineering," by Balling, R.J., Pister, K.S. and Polak, E.,December 1981, (PB82 218496)A07.
UCB/EERC-81/20 "Optimal Design of Seismic-Resistant Planar Steel Frames: by Balling, R.J.. Ciampi, V. and Pister, K.S., December 1981, (PB82 220I79)A07.
UCB/EERC-82101 "Dynamic Behavior of Ground for Seismic Analysis of Lifeline Systems: by Sato, T. and Der Kiureghian, A., January 1982, (PB82 218926)A05.
UCB/EERC-82102 "Shaking Table Tests of a Tubular Steel Frame Model; by Ghanaat, Y. and Cough, R.W., January 1982, (PB82 220 161)A07.
UCB/EERC-82/03 -Behavior of a Piping System under Seismic Excitation: Experimental Investigations of a Spatial Piping System supported by Mechanical Shock Arrestors," by Schneider,S., Lee, H.-M. and Godden, W. G., May 1982, (PB83 172 544)A09.
UCB/EERC-82/04 "New Approaches for the Dynamic Analysis of Large Structural Systems: by Wilson, E.L., June 1982, (PB83 148 080)AOS.
UCB/EERC-82105 "Model Study of Effects of Damage on the Vibration Properties of Steel Offshore Platforms: by Shahrivar, F. and Bouwkamp, J.G.,June 1982, (PB83 148 742)AIO.
UCB/EERC-82106 "States of the Art and Pratice in the Optimum Seismic Design and Analytical Response Prediction of RIC Frame Wall Structures; byAktan, A.E. and Benero, V.V., July 1982, (PB83 147 736)A05.
UCB/EERC-82107 "Further Study of the Earthquake Response of a Broad Cylindrical Liquid-Storage Tank Model: by Manos, G.c. and Gough; R.W.,July 1982. (PB83 147 744)A I!.
UCB/EERC-82108 "An Evaluation of the Design and Analytical Seismic Response of a Seven Story Reinforced Concrete Frame," by Charney, F.A. andBertero. V.V., July 1982, (pB83 157 628)A09.
UCB/EERC-82/09 "Fluid-Structure Interactions: Added Mass Computations for Incompressible Fluid: by Kuo, J.S.-H., August 1982, (PB83 156 281)A07.
UCB/EERC-82/l0 'Joint-Dpening Nonlinear Mechanism: Interface Smeared Crack Model: by Kuo, J.S.-H., August 1982, (PB83 149 195)A05.
123
UCB/EERC-82/11 "Dynamic Response Analysis of Techi Dam," by Clough, R.W., Stephen, RM. and Kuo, J.S.-H., August 1982, (PB83 147 496)A06.
UCB/EERC-82/12 "Prediction of the Seismic Response of RIC Frame-Coupled Wall Structures," by Aktan, A.E., Bertero, V.V. and Piazzo, M., August1982, (PB83 149 203)A09.
UCB/EERC-82/13 -Preliminary Report on the Smart I Strong Motion Array in Taiwan," by Bolt, B.A., Loh, C.H., Penzien, J. and Tsai, Y.B., August1982, (PB83 159 400)AIO.
UCB/EERC"82114 "Seismic Behavior of an EccentricalIy X-Braced Steel Structure," by Yang, M.S., September 1982, (PB83 260 778)AI2.
UCB/EERC-82/15 "The Performance of Stairways in Earthquakes: by Roha, C., Axley, J.W. and Bertero, V.V., September 1982, (PB83 157 693)A07.
UCB/EERC-82/16 "The Behavior of Submerged Multiple Bodies in Earthquakes: by Liao, W.-G., September 1982, (PB83 158 709)A07.
UCB/EERC-82/l7 "Effects of Concrete Types and Loading Conditions on Local Bond"Slip Relationships: by CowelI, A.D., Popov, E.P. and Bertero, V.V.,September 1982, (PB83 153 577)A04.
UCB/EERC.82/18 "Mechanical Behavior of Shear WalI Vertical Boundary Members: An Experimental Investigation: by Wagner, M.T. and Bertero, V.V.,October 1982, (PB83 159 764)A05.
UCB/EERC.82/19 "Experimental Studies of Multi-support Seismic Loading on Piping Systems: by Kelly, J.M. and Cowell, A.D., November 1982, (PB90262684)A07.
UCB/EERC-82/20 "Generalized Plastic Hinge Concepts for 3D Beam·Column Elements," by Chen, P. F.-S. and Powell, G.H., November 1982, (PB83 247981)AI3.
UCB/EERC-82/21 "ANSR-II: General Computer Program for Nonlinear Structural Analysis," by Oughourlian, C.V. and Powell, G.H., November 1982,(PB83 251 330)AI2.
UCB/EERC·82122 "Solution Strategies for StaticalIy Loaded Nonlinear Structures: by Simons, J.W. and Powell, G.H., November 1982, (PB83 197970)A06.
UCB/EERC-82/23 "Analytical Model of Deformed Bar Anchorages under Generalized Excitations: by Ciampi, V., Eligehausen, R., Bertero, V.V. andPopov, E.P., November 1982, (PB83 169 532)A06.
UCB/EERC"82/24 "A Mathematical Model for the Response of Masonry Walls to Dynamic Excitations: by Sucuoglu, H., Mengi, Y. and McNiven, RD.,November 1982, (PB83 169 OII)A07.
UCB/EERC-82/25 "Earthquake Response Considerations of Broad Liquid Storage Tanks: by Cambra, F.J., November 1982, (PB83 251 215)A09.
UCB/EERC·82/26 "Computational Models for Cyclic Plasticity, Rate Dependence and Creep," by Mosaddad, B. and Powell, G.H., November 1982, (PB83245 829)A08.
UCB/EERC·82127 "Inelastic Analysis of Piping and Tubular Structures: by Mahasuverachai, M. and Powell, G.H., November 1982, (PB83 249 987)A07.
UCB/EERC-83/0l "The Economic Feasibility of Seismic Rehabilitation of Buildings by Base Isolation: by Kelly, J.M., January 1983, (PB83 197 988)A05.
UCB/EERC-83/02 "Seismic Moment Connections for Moment-Resisting Steel Frames.: by Popov, E.P., January 1983, (PB83 195 412)A04.
UCB/EERC-83/03 -Design of Links and Beam-ta-Column Connections for Eccentrically Braced Steel Frames: by Popov, E.P. and Malley, J.O., January1983, (PB83 194 811 )A04.
UCB/EERC·83/04 "Numerical Techniques for the Evaluation of Soil·Structure Interaction Effects in the Time Domain: by Bayo, E. and Wilson, E.L.,February 1983, (PB83 245 605)A09.
UCB/EERC-83/05 "A Transducer for Measuring the Internal Forces in the Columns of a Frame-Wall Reinforced Concr::te Structure: by Sause, R. andBertero, V.V., May 1983, (PB84 119 494)A06.
UCB/EERC·83/06 "Dynamic Interactions Between Floating Ice and Offshore Structures," by Croteau, P., May 1983, (PB84 119 486)A16.
UCB/EERC·83/07 'Dynamic Analysis of Multiply Tuned and Arbitrarily Supported Secondary Systems," by Igusa, T. and Der Kiureghian, A., July 1983,(PB84 118 272)AII.
UCB/EERC·83/08 "A Laboratory Study of Submerged Multi·body Systems in Earthquakes: by Ansari, G.R., June 1983, (PB83 261 842)AI7.
UCB/EERC-83/09 "Effects of Transient Foundation Uplift on Earthquake Response of Structures: by Vim, C.-S. and Chopra, A.K., June 1983, (PB83 261396)A07.
UCB/EERC-83/10 'Optimal Design of Friction·Braced Frames under Seismic Loading," by Austin, M.A. and Pister, K.S., June 1983, (PB84 119 288)A06.
UCB/EERC·83/11 'Shaking Table Study of Single-Story Masonry Houses: Dynamic Performance under Three Component Seismic Input and Recommendations: by Manos, G.c., Clough, R.W. and Mayes, RL., July 1983, (UCB/EERC·83/11)A08.
UCB/EERC·83/12 -Experimental Error Propagation in Pseudodynamic Testing," by Shiing, P.B. and Mahin, S.A., June 1983, (PB84 119 270)A09.
UCB/EERC·83/13 "Experimental and Analytical Predictions of the Mechanical Characteristics of a 1/5"scale Model of a 7"story RIC Frame-Wall BuildingStructure: by Aktan, A.E., Bertero, V.V., Chowdhury, A.A. and Nagashima, T., June 1983, (PB84 119 213)A07.
UCB/EERC"83/14 "Shaking Table Tests of Large-Panel Precast Concrete Building System Assemblages: by Oliva, M.G. and Oough, RW., June 1983,(PB86 110 210/AS)AII.
UCB/EERC·83/15 "Seismic Behavior of Active Beam Links in Eccentrically Braced Frames: by Hjelmstad, K.D. and Popov, E.P., July 1983, (PB84 119676)A09.
UCB/EERC"83/16 "System Identification of Structures with Joint Rotation," by Dimsdale, J.S., July 1983, (PB84 192 21O)A06.
UCB/EERC"83/17 "Construction of Inelastic Response Spectra for Single.Degree-of-Freedom Systems: by Mahin, S. and Lin, J., June 1983, (PB84 208834)A05.
UCBIEERC·83/18 "Interactive Computer Analysis Methods for Predicting the Inelastic Cyclic Behaviour of Structural Sections: by Kaba, S. and Mahin,S., July 1983, (PB84 192 012)A06.
UCB/EERC·83/19 "Effects of Bond Deterioration on Hysteretic Behavior of Reinforced Concrete Joints: by Filippou, F.C., Popov, E.P. and Bertero, V.V.,August 1983, (PB84 192 020)AIO.
124
UCB/EERC-83/20 "Correlation of Analytical and Experimental Responses of Large-Panel Precast Building Systems; by Oliva, M.G., Clough, R.W., Velkov, M. and Gavrilovic, P., May 1988, (PB90 262 692)A06.
UCB/EERC-83121 'Mechanical Characteristics of Materials Used in a 115 Scale Model of.a 7-Story Reinforced Concrete Test Structure; by Bertero, V.V.,Aktan, A.E., Harris, H.G. and Chowdhury, A.A., October 1983, (PB84 193 697)A05.
UCB/EERC·83122 'Hybrid Modelling of Soil-Structure Interaction in Layered Media; by Tzong, T.-J. and Penzien, J., October 1983, (PB84 192 178)A08.
UCB/EERC-83123 'Local Bond Stress-Slip Relationships of Deformed Bars under Generalized Excitations,' by Eligehausen, R., Popov, E.P. and Bertero,V.V., October 1983, (PB84 192 848)A09.
UCB/EERC·83124 'Design Considerations for Shear Links in Eccentrically Braced Frames; by Malley, J.O. and Popov, E.P., November 1983, (PB84 192I86)A07.
UCB/EERC-84/01 'Pseudodynamic Test Method for Seismic Performance Evaluation: Theory and Implementation; by Shing, P.-S.B. and Mahin, S.A.,January 1984, (PB84 190 644)A08.
UCB/EERC-84/02 'Dynamic Response Behavior of Kiang Hong Dian Dam; by Clough, R.W., Chang, K.-T., Chen, H.-Q. and Stephen, R.M., April 1984,(PB84 209 402)A08.
UCB/EERC-84/03 'Refined Modelling of Reinforced Concrete Columns for Seismic Analysis; by Kaba, SA and Mahin, SA, April 1984, (PB84 234384)A06.
UCB/EERC-84/04 •A New Floor Response Spectrum Method for Seismic Analysis of Multiply Supported Secondary Systems; by Asfura, A. and DerKiureghian, A., June 1984, (PB84 239 417)A06.
UCB/EERC-84/05 'Earthquake Simulation Tests and Associated Studies of a 1/5th-scale Model of a 7-Story RIC Frame-Wall Test Structure,' by Bertero,V.V., Aktan, A.E., Charney, EA. and Sause, R., June 1984, (PB84 239 409)A09.
UCB/EERC-84/06 "RIC Structural Walls: Seismic Design for Shear; by Aktan, A.E. and Bertero, V.V., 1984.
UCB/EERC-84/07 'Behavior of Interior and Exterior Flat-Plate Connections subjected to Inelastic Load Reversals,' by Zee, H.L. and Moehle, J.P., August1984, (PB86 117 629/AS)A07.
UCB/EERC-84/08 'Experimental Study of the Seismic Behavior of a TWO-Story Flat-Plate Structure,' by Moehle, J.P. and Diebold, J.W., August 1984,(PB86 122 553/AS)AI2.
UCB/EERC-84/09 'Phenomenological Modeling of Steel Braces under Cyclic Loading; by Ikeda, K., Mahin, S.A. and Dermitzakis, S.N., May 1984, (PB86132 198/AS)A08.
UCB/EERC-84/10 'EarthQuake Analysis and Response of Concrete Gravity Dams; by Fenves, G. and Chopra, A.K., August 1984, (PB85 193902lAS)A11.
UCB/EERC-841l1 -EAGD-84: A Computer Program for Earthquake Analysis of Concrete Gravity Dams; by Fenves, G. and Chopra, A.K., August 1984,(PB85 193 613/AS)A05.
UCB/EERC-84/12 'A Refined Physical Theory Model for Predicting the Seismic Behavior of Braced Steel Frames; by Ikeda, K. and Mahin, S.A., July1984, (PB85 191 450/AS)A09.
UCB/EERC-84/13 -Earthquake Engineering Research at Berkeley - 1984; by, August 1984, (PB85 197 341/AS)AIO.
UCB/EERC-84/14 'Moduli and Damping Factors for Dynamic Analyses of Cohesionless Soils; by Seed, H.B., Wong, R.T., Idriss, LM. and Tokimatsu, K.,September 1984, (PB85 191 468/AS)A04.
UCB/EERC-84/15 -The Influence of SPT Procedures in Soil Liquefaction Resistance Evaluations; by Seed, H.B., Tokimatsu, K., Harder, L.E and Chung,R.M., October 1984, (PB85 191 7321AS)A04.
UCB/EERC-841l6 'Simplified Procedures for the Evaluation of Settlements in Sands Due to Earthquake Shaking; by Tokimatsu, K. and Seed, H.B.,October 1984, (PB85 197 887/AS)A03.
UCB/EERC-84/17 'Evaluation of Energy Absorption Characteristics of Highway Bridges Under Seismic Conditions - Volume I (PB90 262 627)A16 andVolume II (Appendices) (PB90 262 635)AI3; by Imbsen, R.A. and Penzien, J., September 1986.
UCB/EERC-84/18 'Structure-Foundation Interactions under Dynamic Loads; by Liu, W.D. and Penzien, J., November 1984, (PB87 124 889/AS)AII.
UCB/EERC-84/19 'Seismic Modelling of Deep Foundations; by Chen. e.-H. and Penzien, J., November 1984, (PB87 124 798/AS)A07.
UCB/EERC-84/20 'Dynamic Response Behavior of Quan Shui Dam; by Clough, R.W., Chang, K.-T., Chen, H.-Q., Stephen, R.M., Ghanaat, Y. and Qi,J.-H., November 1984, (PB86 115177/AS)A07.
UCB/EERC-85/0 I 'Simplified Methods of Analysis for Earthquake Resistant Design of Buildings; by Cruz, E.F. and Chopra, A.K., February 1985, (PB86112299/AS)AI2.
UCB/EERC-85/02 'Estimation of Seismic Wave Coherency and Rupture Velocity using the SMART I Strong-Motion Array Recordings; by Abrahamson,N.A., March 1985, (PB86 214 343)A07.
UCB/EERC-85/03 -Dynamic Properties of a Thirty Story Condominium Tower Building; by Stephen, R.M., Wilson, E.L. and Stander, N., April 1985,(PB86 118965/AS)A06.
UCB/EERC-85/04 'Development of Substructuring Techniques for On·Line Computer Controlled Seismic Performance Testing," by Dermitzakis, S. andMahin, S., February 1985, (PB86 13294I1AS)A08.
UCB/EERC-85/05 •A Simple Model for Reinforcing Bar Anchorages under Cyclic Excitations,' by Filippou, F.e., March 1985, (PB86 112 919/AS)A05.
UCB/EERC-85/06 'Racking Behavior of Wood-framed Gypsum Panels under Dynamic Load; by Oliva, M.G., June 1985, (PB90 262 643)A04.
UCB/EERC-85/07 'Earthquake Analysis and Response of Concrete Arch Dams; by Fok, K.-L. and Chopra, A.K., June 1985, (PB86 1396721AS)AIO.
UCB/EERC-85/08 "Effect of Inelastic Behavior on the Analysis and Design of Earthquake Resistant Structures; by Lin, J.P. and Mahin, S.A., June 1985,(PB86 135340/AS)A08.
UCB/EERC-85/09 'Earthquake Simulator Testing of a Base-Isolated Bridge Deck; by Kelly, J.M.• Buckle, LG. and Tsai, H.-C., January 1986. (PB87 1241521AS)A06.
125
UCB/EERC-85/10 "Simplified Analysis for Earthquake Resistant Design of Concrete Gravity Dams: by Fenves, G. and Chopra, A.K., June 1986, (PB87124 160/AS)A08.
UCB/EERC-85/11 "Dynamic Interaction Effects in Arch Dams: by Clough, R.W., Chang, K.-T., Chen, H.-Q. and Ghanaat, Y., October 1985, (PB86135027/AS)A05.
UCB/EERC-85/12 "Dynamic Response of Long Valley Dam in the Mammoth Lake Earthquake Series of May 25-27, 1980: by Lai, S. and Seed, H.B.,November 1985, (PB86 142304/AS)A05.
UCB/EERC-85113 -A Methodology for Computer-Aided Design of Earthquake-Resistant Steel Structures: by Austin, M.A., Pister, K.S. and Mahin, S.A.,December 1985, (PB86 159480/AS)AlO .
UCB/EERC"85114 "Response of Tension-Leg Platforms to Vertical Seismic Excitations: by Liou, G.-S., Penzien, J. and Yeung, R.W., December 1985,(PB87 124 8711AS)A08.
UCB/EERC-85/1S -Cyclic Loading Tests of Masonry Single Piers: Volume 4 - Additional Tests with Height to Width Ratio of 1: by Sveinsson, B.,McNiven, H.D. and Sucuoglu, H., December 1985.
UCB/EERC-85/16 -An Experimental Program for Studying the Dynamic Response of a Steel Frame with a Variety of Inlill Partitions: by Yanev, B. andMcNiven, H.D., December 1985, (PB90 262 676)AOS.
UCB/EERC-86/01 -A Study of Seismically Resistant Eccentrically Braced Steel Frame Systems: by Kasai, K. and Popov, E.P., January 1986, (PB87 124178/AS)AI4.
UCB/EERC-86/02 -Design Problems in Soil Liquefaction: by Seed, H.B., February 1986, (PB87 124 186/AS)A03.
UCB/EERC-86/03 -Implications of Recent Earthquakes and Research on Earthquake-Resistant Design and Construction of Buildings: by Bertero, V.V.,March 1986, (PB87 124 194/AS)A05.
UCB/EERC-86/04 -The Use of Load Dependent Vectors for Dynamic and Earthquake Analyses: by Leger, P., Wilson, E.L. and Clough, R.W., March1986, (PB87 124 202/AS)AI2.
UCB/EERC-86/0S -Two Beam-To-Column Web Connections: by Tsai, K.-c. and Popov, E.P., April 1986, (PB87 124 301lAS)A04.
UCB/EERC-86/06 -Determination of Penetration Resistance for Coarse-Grained Soils using the Becker Hammer Drill: by Harder, L.F. and Seed, H.B.,May 1986, (PB87 124 210/AS)A07.
UCB/EERC-86/07
UCB/EERC-87/03
UCB/EERC-86/10
UCB/EERC-86111
UCB/EERC-86/08
UCB/EERC-86/09
"A Mathematical Model for Predicting the Nonlinear Response of Unreinforced Masonry Walls to In-Plane Earthquake Excitations,' byMengi, Y. and McNiven, H.D., May 1986, (PB87 124 780/AS)A06.
-The 19 September 1985 Mexico Earthquake: Building Behavior: by Bertero, V.V., July 1986.
-EACD-3D: A Computer Program for Three-Dimensional Earthquake Analysis of Concrete Dams: by Fok, K.-L., Hall, J.F. andChopra, A.K., July 1986, (PB87 124 228/AS)A08.
-Earthquake Simulation Tests and Associated Studies of a 0.3-Scale Model of a Six-Story Concentrically Braced Steel Structure: byUang, e,-M. and Bertero, V.V., December 1986, (PB87 163 S64/AS)AI7.
-Mechanical Characteristics of Base Isolation Bearings for a Bridge Deck Model Test: by Kelly, J.M., Buckle, I.G. and Koh, c.-G.,November 1987, (PB90 262 668)A04.
"Effects of Axial Load on Elastomeric Isolation Bearings: by Koh, C.-G. and Kelly, J.M., November 1987.
"The FPS Earthquake Resisting System: Experimental Report," by Zayas, V.A., Low, S.S. and Mahin, S.A., June 1987.
-Earthquake Simulator Tests and Associated Studies of a O.3-Scale Model of a Six-Story Eccentrically Braced Steel Structure,- by Whittaker, A., Uang, C.-M. and Bertero, V.V., July 1987.
"A Displacement Control and Uplift Restraint Device for Base-Isolated Structures: by Kelly, J.M., Griffith, M.C. and Aiken, I.D., April1987.
UCB/EERC-87/04 "Earthquake Simulator Testing of a Combined Sliding Bearing and Rubber Bearing Isolation System: by Kelly, J.M. and Chalhoub,M.S., 1987.
UCB/EERC-86/12
VCB/EERC-87/01
UCB/EERC-87/02
UCB/EERC-87/0S "Three-Dimensional Inelastic Analysis of Reinforced Concrete Frame-Wall Structures: by Moazzami, S. and Benero, V.V., May 1987.
UCB/EERC-87/06 "Experiments on Eccentrically Braced Frames with Composite Floors: by Rides, J. and Popov, E., June 1987.
UCB/EERC"87/07 "Dynamic Analysis of Seismically Resistant Eccentrically Braced Frames: by Ricles, J. and Popov, E., June 1987.
UCB/EERC-87/08 "Undrained Cyclic Triaxial Testing of Gravels-The Effect of Membrane Compliance," by Evans, M.D. and Seed, H.B., July 1987.
UCB/EERC-87/09 "Hybrid Solution Techniques for Generalized Pseudo-Dynamic Testing," by Thewalt, C. and Mahin, S.A., July 1987.
UCB/EERC-87/10 "Ultimate Behavior of Butt Welded Splices in Heavy Rolled Steel Sections: by Bruneau, M., Mahin, S.A. and Popov, E.P., September1987.
UCB/EERC-87/I1 "Residual Strength of Sand from Dam Failures in the Chilean Earthquake of March 3, 1985: by De Alba, P., Seed, H.B., Retamal, E.and Seed, R.B., September 1987.
UCB/EERC-87112 "Inelastic Seismic Response of Structures with Mass or Stiffness Eccentricities in Plan: by Bruneau, M. and Mahin, S.A., September1987, (PB90 262 650)AI4.
UCB/EERC-87113 "CSTRUCT: An Interactive Computer Environment for the Design and Analysis of Earthquake Resistant Steel Structures," by Austin,M.A., Mahin, S.A. and Pister, K.S., September 1987.
UCB/EERC-87/14 -Experimental Study of Reinforced Concrete Columns Subjected to Multi-Axial Loading: by Low, S.S. and Moehle, J.P., September1987.
UCB/EERC-87/IS "Relationships between Soil Conditions and Earthquake Ground Motions in Mexico City in the Earthquake of Sept. 19, 1985: by Seed,H.B., Romo, M.P., Sun, J., Jaime, A. and Lysmer, J., October 1987.
UCB/EERC-87/16 "Experimental Study of Seismic Response of R. C. Setback Buildings: by Shahrooz, B.M. and Moehle, J.P., October 1987.
126
UCB/EERC-88/03
UCB/EERC-88/04
UCB/EERC-87/17
UCB/EERC-87/18
UCB/EERC-87/19
UCB/EERC-88/06
UCB/EERC-88/07
UCB/EERC-88/08
UCB/EERC-88/09
UCB/EERC-87/20
UCB/EERC"87/21
UCB/EERC-87122
UCB/EERC-88/0 I
UCB/EERC-88/02
"The Effect of Slabs on the Flexural Behavior of Beams: by Pantazopoulou, S.J. and Moehle, J.P., October 1987, (PB90 262 7oo)A07.
"Design Procedure for R-FBI Bearings: by Mostaghel, N. and Kelly, J.~., November 1987, (PB90 262 718)A04.
"Analytical Models for Predicting the Lateral Response of R C Shear Walls: Evaluation of their Reliability," by Vulcano, A. and Bertero, V.V., November 1987.
"Earthquake Response of Torsionally-Coupled Buildings: by Hejal, R. and Chopra, A.K., December 1987.
"Dynamic Reservoir Interaction with Monticello Dam: by Clough, R.W., Ghanaat, Y. and Qiu, X-F., December 1987.
"Strength Evaluation of Coarse-Grained Soils: by Siddiqi, F.H., Seed, R.B., Chan, C.K., Seed, H.B. and Pyke, R.M., December 1987.
"Seismic Behavior of Concentrically Braced Steel Frames: by Khatib, I., Mahin, S.A. and Pister, K.S., January 1988.
"Experimental Evaluation of Seismic Isolation of Medium-Rise Structures Subject to Uplift: by Griffith, M.C., Kelly, J.M., Coveney,VA and Koh, e.G., January 1988.
"Cyclic Behavior of Steel Double Angle Connections," by Astaneh"AsI, A. and Nader, M.N., January 1988.
"Re-evaluation of the Slide in the Lower San Fernando Dam in the Earthquake of Feb. 9, 1971: by Seed, H.B., Seed, R.B., Harder,L.F. and Jong, H.·L., April 1988.
UCB/EERC"88/05 -Experimental Evaluation of Seismic Isolation of a Nine-Story Braced Steel Frame Subject to Uplift," by Griffith, M.e., Kelly, J.M. andAiken, I.D., May 1988.
"DRAIN-2DX User Guide.," by Allahabadi, R. and Powell, G.H., March 1988.
"Cylindrical Fluid Containers in Base-Isolated Structures: by Chalhoub, M.S. and Kelly, J.M. , April 1988.
"Analysis of Near·Source Waves: Separation of Wave Types using Strong Motion Array Recordings: by Darragh, R.B., June 1988.
"Alternatives to Standard Mode Superposition for Analysis of Non-Classically Damped Systems: by Kusainov, A.A. and Clough, R.W.,June 1988.
UCB/EERC-88/10 'The Landslide at the Port of Nice on October 16, 1979: by Seed, H.B., Seed, R.B., Schlosser, F., Blondeau, F. and Juran, I., June1988.
UCB/EERC-88/11
UCB/EERC-88/12
UCB/EERC-881l3
UCB/EERC-88/14
UCB/EERC-88/15
UCB/EERC-881l6
UCB/EERC·88/17
UCB/EERC-88/18
UCB/EERC-88/19
UCB/EERC-88/20
UCB/EERC-89/01
UCB/EERC-89/02
UCB/EERC-89/03
UCB/EERC-89/04
UCB/EERC-89/05
UCB/EERC-89/06
UCB/EERC-89/07
UCB/EERC-89/08
UCB/EERC-89/09
UCB/EERC·89/10
UCB/EERC-891l1
UCB/EERC-89/12
UCB/EERC-89/13
UCB/EERC-89/14
UCB/EERC-891l5
"Liquefaction Potential of Sand Deposits Under Low Levels of Excitation: by Carter, D.P. and Seed, H.B., August 1988.
"Nonlinear Analysis of Reinforced Concrete Frames Under Cyclic Load Reversals, - by Filippou, F.e. and Issa, A., September 1988.
"Implications of Recorded Earthquake Ground Motions on Seismic Design of Building Structures: by Uang, e.-M. and Bertero, V.V.,November 1988.
"An Experimental Study of the Behavior of Dual Steel Systems," by Whittaker, A.S. , Uang, C.-M. and Bertero, V.V., September 1988.
"Dynamic Moduli and Damping Ratios for Cohesive Soils: by Sun, J.I., Golesorkhi, R. and Seed, H.B., August 1988.
"Reinforced Concrete Flat Plates Under Lateral Load: An Experimental Study Including Biaxial Effects: by Pan, A. and Moehle, J.,October 1988.
"Earthquake Engineering Research at Berkeley - 1988: by EERC, November 1988.
"Use of Energy as a Design Criterion in Earthquake·Resistant Design: by Uang, e.-M. and Bertero, V.V., November 1988.
"Steel Beam-Column Joints in Seismic Moment Resisting Frames," by Tsai, K.-c. and Popov, E.P., November 1988.
"Base Isolation in Japan, 1988: by Kelly, J.M., December 1988.
-Behavior of Long Links in Eccentrically Braced Frames,- by Engelhardt, M.D. and Popov, E.P., January 1989.
"Earthquake Simulator Testing of Steel Plate Added Damping and Stiffness Elements: by Whittaker, A., Bertero, V.V., Alonso, J. andThompson, e., January 1989.
"Implications of Site Effects in the Mexico City Earthquake of Sept. 19, 1985 for Earthquake-Resistant Design Criteria in the San Francisco Bay Area of California: by Seed, H.B. and Sun, J.I., March 1989.
-Earthquake Analysis and Response of Intake-outlet Towers: by Goyal, A. and Chopra, A.K., July 1989.
"The 1985 Chile Earthquake: An Evaluation of Structural Requirements for Bearing Wall Buildings: by Wallace, J.W. and Moehle,J.P., July 1989.
"Effects of Spatial Variation of Ground Motions on Large Multiply-Supported Structures: by Hao, H., July 1989.
-EADAP· Enhanced Arch Dam Analysis Program: Users's Manual: by Ghanaat, Y. and Clough, R.W., August 1989.
"Seismic Performance of Steel Moment Frames Plastically Designed by Least Squares Stress Fields,- by Ohi, K. and Mahin, S.A.,August 1989.
-Feasibility and Performance Studies on Improving the Earthquake Resistance of New and Existing Buildings Using the Friction Pendulum System: by zayas, V., Low, S., Mahin, SA. and Bozzo, L., July 1989.
"Measurement and Elimination of Membrane Compliance Effects in Undrained Triaxial Testing: by Nicholson, P.G., Seed. R.B. andAnwar, H., September 1989.
'Static Tilt Behavior of Unanchored Cylindrical Tanks: by Lau, D.T. and Clough, R.W., September 1989.
"ADAP-88: A Computer Program for Nonlinear Earthquake Analysis of Concrete Arch Dams: by Fenves, G.L., Mojtahedi, S. and Rei"mer, R.B., September 1989.
'Mechanics of Low Shape Factor Elastomeric Seismic Isolation Bearings,' by Aiken, 1.0., Kelly, J.M. and Tajirian, F., December 1989.
'Preliminary Report on the Seismological and Engineering Aspects of the October 17, 1989 Santa Cruz (Loma Prieta) Earthquake: byEERC, October 1989.
-Experimental Studies of a Single Story Steel Structure Tested with Fixed, Semi·Rigid and Flexible Connections: by Nader. M.N. andAstaneh-AsI, A., August 1989.
127
UCBlEERC-89/16
UCB/EERC-90/0 I
UCB/EERC-90/02
UCB/EERC-90/03
UCB/EERC-90/04
UCB/EERC-90/05
UCB/EERC-90/06
UCB/EERC-90/07
UCB/EERC-90/08
UCB/EERC-90/09
UCB/EERC-90/10
UCB/EERC-90/11
UCB/EERC-90112
UCB/EERC-90/13
UCB/EERC-90/14
UCB/EERC-90/15
UCB/EERC-90/16
UCB/EERC-90/17
UCB/EERC-90118
UCB/EERC-90/19
UCB/EERC-90/20
UCB/EERC-90/21
UCB/EERC-91/01
UCB/EERC-91/02
UCB/EERC-91/03
UCB/EERC-91/04
UCB/EERC-91105
UCB/EERC-91/06
-Collapse of the Cypress Street Viaduct as a Result of the Loma Prieta Earthquake: by Nims, D.K., Miranda, E., Aiken, I.D.. Whittaker, A.S. and Bertero, V.V., November 1989.
-Mechanics of High-Shape Factor Elastomeric Seismic Isolation Bearings," by Kelly, J.M., Aiken, LD. and Tajirian, F.F., March 1990.
"Javid's Paradox: The Influence of Preform on the Modes of Vibrating Beams: by Kelly, J.M., Sackman, J.L. and Javid, A., May 1990.
-Earthquake Simulator Testing and Analytical Studies of Two Energy-Absorbing Systems for Multistory Structures," by Aiken, LD. andKelly, J.M., October 1990.
"Damage to the San Francisco-Oakland Bay Bridge During the October 17,1989 Earthquake: by Astaneh, A., June 1990.
"Preliminary Report on the Principal Geotechnical Aspects of the October 17, 1989 Loma Prieta Earthquake," by Seed, R.B., Dickenson, S.E., Riemer, M.F., Bray, J.D., Sitar, N., Mitchell, J.K., Idriss, LM., Kayen, R.E., Kropp, A., Harder, L.F., Jr. and Power, M.S.,April 1990.
"Models of Critical Regions in Reinforced Concrete Frames Under Seismic Excitations: by Zulfiqar, N. and Filippou, F., May 1990.
-A Unified Earthquake-Resistant Design Method for Steel Frames Using ARMA Models: by Takewaici, I., Conte, J.P., Mahin, S.A. andPister, K.S., June 1990.
"Soil Conditions and Earthquake Hazard Mitigation in the Marina District of San Francisco: by Mitchell, J.K., Masood, T., Kayen,R.E. and Seed, R.B., May 1990.
"Influence of the Earthquake Ground Motion Process and Structural Properties on Response Characteristics of Simple Structures: byConte, J.P., Pister, K.S. and Mahin, S.A., July 1990.
"Experimental Testing of the Resilient-Friction Base Isolation System: by Clark, P.W. and Kelly, J.M., July 1990.
-Seismic Hazard Analysis: Improved Models, Uncertainties and Sensitivities," by Araya, R. and Der Kiureghian, A., March 1988.
"Effects of Torsion on the Linear and Nonlinear Seismic Response of Structures: by Sedarat, H. and Bertero, V.V., September 1989.
-The Effects of Tectonic Movements on Stresses and Deformations in Earth Embankments: by Bray, J. D., Seed, R. B. and Seed, H. B.,September 1989.
-Inelastic Seismic Response of One-Story, Asymmetric-Plan Systems: by Geel, R.K. and Chopra, A.K., October 1990.
-Dynamic Crack Propagation: A Model for Near-Field Ground Motion.: by Seyyedian, H. and Kelly, J.M., 1990.
"SenSitivity of Long-Period Response Spectra to System Initial Conditions: by Blasquez, R., Ventura, C. and Kelly, J.M., 1990.
"Behavior of Peak Values and Spectral Ordinates of Near-Source Strong Ground-Motion over a Dense Array: by Niazi, M., June 1990.
"Material Characterization of Elastomers used in Earthquake Base Isolation: by Papoulia, K.D. and Kelly, J.M., 1990.
-Cyclic Behavior of Steel Top-and-Bottom Plate Moment Connections: by Harriott, J.D. and Astaneh, A., August 1990.
"Seismic Response Evaluation of an Instrumented Six Story Steel Building," by Shen, J.-H. and Astaneh, A., December 1990.
"Observations and Implications of Tests on the Cypress Street Viaduct Test Structure: by Mahin, S.A., Moehle, J.P., Stephen. R.M.,Bolio. M. and Qi, X., December 1990.
-Experimental Evaluation of Nitinol for Energy Dissipation in Structures: by Nims, D.K., Sasaki, K.K. and Kelly, J.M., 1991.
"Displacement Design Approach for Reinforced Concrete Structures Subjected to Earthquakes: by Qi, X. and Moehle, J.P.. January1991.
-Shake Table Tests of Long Period Isolation System for Nuclear Facilities at Soft Soil Sites, by Kelly, J.M., March 1991.
-Dynamic and Failure Characteristics of Bridgestone Isolation Bearings: by Kelly, J.M., April 1991.
"Base Sliding Response of Concrete Gravity Dams to Earthquakes: by Chopra, A.K. and Zhang, L.. May 1991.
"Computation of Spatially Varying Ground Motion and Foundation-Rock Impedance Matrices for Seismic Analysis of Arch Dams: byZhang, L. and Chopra, A.K., May 1991.
128